[v2,1/1] tricore: fixed faulty conditions for extr and imask

From: Andreas Konopik <andreas.konopik@efs-auto.de>

Signed-off-by: Andreas Konopik <andreas.konopik@efs-auto.de>
Signed-off-by: Georg Hofstetter <georg.hofstetter@efs-auto.de>
Signed-off-by: David Brenken <david.brenken@efs-auto.de>
---
 target/tricore/translate.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

Hi David and Andreas,

On 2/11/21 12:53 PM, David Brenken wrote:
> From: Andreas Konopik <andreas.konopik@efs-auto.de>

Here is a good place to explain why you need this change,
how did you noticed it (example of opcode and conditions
reaching this issue) - eventually provide a reproducer
(asm dump could be enough) - and also eventually a reference
to the manual (chapter, table) justifying your change.

See also:
https://tbaggery.com/2008/04/19/a-note-about-git-commit-messages.html
https://chris.beams.io/posts/git-commit/#why-not-how

> 
> Signed-off-by: Andreas Konopik <andreas.konopik@efs-auto.de>
> Signed-off-by: Georg Hofstetter <georg.hofstetter@efs-auto.de>
> Signed-off-by: David Brenken <david.brenken@efs-auto.de>
> ---
>  target/tricore/translate.c | 8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/target/tricore/translate.c b/target/tricore/translate.c
> index 7752630ac1..ebeddf8f4a 100644
> --- a/target/tricore/translate.c
> +++ b/target/tricore/translate.c
> @@ -5777,8 +5777,8 @@ static void decode_rcpw_insert(DisasContext *ctx)
>      switch (op2) {
>      case OPC2_32_RCPW_IMASK:
>          CHECK_REG_PAIR(r2);
> -        /* if pos + width > 31 undefined result */
> -        if (pos + width <= 31) {
> +        /* if pos + width > 32 undefined result */
> +        if (pos + width <= 32) {
>              tcg_gen_movi_tl(cpu_gpr_d[r2+1], ((1u << width) - 1) << pos);
>              tcg_gen_movi_tl(cpu_gpr_d[r2], (const4 << pos));
>          }
> @@ -6999,7 +6999,7 @@ static void decode_rrpw_extract_insert(DisasContext *ctx)
>  
>      switch (op2) {
>      case OPC2_32_RRPW_EXTR:
> -        if (pos + width <= 31) {
> +        if (pos + width <= 32) {
>              /* optimize special cases */
>              if ((pos == 0) && (width == 8)) {
>                  tcg_gen_ext8s_tl(cpu_gpr_d[r3], cpu_gpr_d[r1]);
> @@ -7021,7 +7021,7 @@ static void decode_rrpw_extract_insert(DisasContext *ctx)
>          break;
>      case OPC2_32_RRPW_IMASK:
>          CHECK_REG_PAIR(r3);
> -        if (pos + width <= 31) {
> +        if (pos + width <= 32) {
>              tcg_gen_movi_tl(cpu_gpr_d[r3+1], ((1u << width) - 1) << pos);
>              tcg_gen_shli_tl(cpu_gpr_d[r3], cpu_gpr_d[r2], pos);
>          }
>

Hi Philippe,

> From: Philippe Mathieu-Daudé <philippe.mathieu.daude@gmail.com> On
> Behalf Of Philippe Mathieu-Daudé
> Sent: Thursday, February 11, 2021 13:10
> Hi David and Andreas,
> 
> On 2/11/21 12:53 PM, David Brenken wrote:
> > From: Andreas Konopik <andreas.konopik@efs-auto.de>
> 
> Here is a good place to explain why you need this change, how did you
> noticed it (example of opcode and conditions reaching this issue) - eventually
> provide a reproducer (asm dump could be enough) - and also eventually a
> reference to the manual (chapter, table) justifying your change.
> 
> See also:
> https://tbaggery.com/2008/04/19/a-note-about-git-commit-messages.html
> https://chris.beams.io/posts/git-commit/#why-not-how

I appreciate your feedback and will write more verbose commit messages in
the future.

According to the TC 1.3.1. Architecture Manual [1; page 174], results are
undefined, if pos + width > 32 or if width = 0.

We found this error because of a different behavior between qemu-tricore
and the real tricore processor. For pos + width = 32, qemu-tricore did not
generate any intermediate code and ran into a different state compared to
the real hardware.

I hope this helps!

> >
> > Signed-off-by: Andreas Konopik <andreas.konopik@efs-auto.de>
> > Signed-off-by: Georg Hofstetter <georg.hofstetter@efs-auto.de>
> > Signed-off-by: David Brenken <david.brenken@efs-auto.de>
> > ---
> >  target/tricore/translate.c | 8 ++++----
> >  1 file changed, 4 insertions(+), 4 deletions(-)
> >
> > diff --git a/target/tricore/translate.c b/target/tricore/translate.c
> > index 7752630ac1..ebeddf8f4a 100644
> > --- a/target/tricore/translate.c
> > +++ b/target/tricore/translate.c
> > @@ -5777,8 +5777,8 @@ static void decode_rcpw_insert(DisasContext
> *ctx)
> >      switch (op2) {
> >      case OPC2_32_RCPW_IMASK:
> >          CHECK_REG_PAIR(r2);
> > -        /* if pos + width > 31 undefined result */
> > -        if (pos + width <= 31) {
> > +        /* if pos + width > 32 undefined result */
> > +        if (pos + width <= 32) {
> >              tcg_gen_movi_tl(cpu_gpr_d[r2+1], ((1u << width) - 1) << pos);
> >              tcg_gen_movi_tl(cpu_gpr_d[r2], (const4 << pos));
> >          }
> > @@ -6999,7 +6999,7 @@ static void
> > decode_rrpw_extract_insert(DisasContext *ctx)
> >
> >      switch (op2) {
> >      case OPC2_32_RRPW_EXTR:
> > -        if (pos + width <= 31) {
> > +        if (pos + width <= 32) {
> >              /* optimize special cases */
> >              if ((pos == 0) && (width == 8)) {
> >                  tcg_gen_ext8s_tl(cpu_gpr_d[r3], cpu_gpr_d[r1]); @@
> > -7021,7 +7021,7 @@ static void decode_rrpw_extract_insert(DisasContext
> *ctx)
> >          break;
> >      case OPC2_32_RRPW_IMASK:
> >          CHECK_REG_PAIR(r3);
> > -        if (pos + width <= 31) {
> > +        if (pos + width <= 32) {
> >              tcg_gen_movi_tl(cpu_gpr_d[r3+1], ((1u << width) - 1) << pos);
> >              tcg_gen_shli_tl(cpu_gpr_d[r3], cpu_gpr_d[r2], pos);
> >          }
> >

[1] https://www.infineon.com/dgdl/tc_v131_instructionset_v138.pdf?fileId=db3a304412b407950112b409b6dd0352

INTERNAL

On Thu, Feb 11, 2021 at 12:53:29PM +0100, David Brenken wrote:
> From: Andreas Konopik <andreas.konopik@efs-auto.de>
> 
> Signed-off-by: Andreas Konopik <andreas.konopik@efs-auto.de>
> Signed-off-by: Georg Hofstetter <georg.hofstetter@efs-auto.de>
> Signed-off-by: David Brenken <david.brenken@efs-auto.de>
> ---
>  target/tricore/translate.c | 8 ++++----
>  1 file changed, 4 insertions(+), 4 deletions(-)

Reviewed-by: Bastian Koppelmann <kbastian@mail.uni-paderborn.de>

Applied to my tricore.next queue. I'll fix the commit message according to
Philippes suggestions. 

Cheers,
Bastian

On Thu, Feb 11, 2021 at 01:49:14PM +0000, Konopik, Andreas (EFS-GH2) wrote:
> Hi Philippe,
> 
> > From: Philippe Mathieu-Daudé <philippe.mathieu.daude@gmail.com> On
> > Behalf Of Philippe Mathieu-Daudé
> > Sent: Thursday, February 11, 2021 13:10
> > Hi David and Andreas,
> > 
> > On 2/11/21 12:53 PM, David Brenken wrote:
> > > From: Andreas Konopik <andreas.konopik@efs-auto.de>
> > 
> > Here is a good place to explain why you need this change, how did you
> > noticed it (example of opcode and conditions reaching this issue) - eventually
> > provide a reproducer (asm dump could be enough) - and also eventually a
> > reference to the manual (chapter, table) justifying your change.
> > 
> > See also:
> > https://tbaggery.com/2008/04/19/a-note-about-git-commit-messages.html
> > https://chris.beams.io/posts/git-commit/#why-not-how
> 
> I appreciate your feedback and will write more verbose commit messages in
> the future.
> 
> According to the TC 1.3.1. Architecture Manual [1; page 174], results are
> undefined, if pos + width > 32 or if width = 0.

While editing the commit message of this patch, I realized that we are not
checking for width=0. A quick test let to:

qemu-system-tricore: ../upstream/tcg/tcg-op.c:217: tcg_gen_sari_i32: Assertion `arg2 >= 0 && arg2 < 32' failed.

As Richard suggested that is not what should happen in case of undefined behaviour.
I'll fix this in a another patch.

Cheers,
Bastian