qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
* [PATCH v3 5/5] target/hexagon: Remove unreachable
  2025-04-07 19:27 [PATCH v3 0/5] misc hexagon patches Brian Cain
@ 2025-04-07 19:27 ` Brian Cain
  2025-04-14 17:19   ` ltaylorsimpson
  0 siblings, 1 reply; 6+ messages in thread
From: Brian Cain @ 2025-04-07 19:27 UTC (permalink / raw)
  To: qemu-devel
  Cc: brian.cain, richard.henderson, philmd, matheus.bernardino, ale,
	anjo, marco.liebel, ltaylorsimpson, alex.bennee, quic_mburton,
	sidneym

We should raise an exception in the event that we encounter a packet
that can't be correctly decoded, not fault.

Signed-off-by: Brian Cain <brian.cain@oss.qualcomm.com>
---
 target/hexagon/decode.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/target/hexagon/decode.c b/target/hexagon/decode.c
index b5ece60450..1db7f1950f 100644
--- a/target/hexagon/decode.c
+++ b/target/hexagon/decode.c
@@ -489,7 +489,6 @@ decode_insns(DisasContext *ctx, Insn *insn, uint32_t encoding)
             insn->iclass = iclass_bits(encoding);
             return 1;
         }
-        g_assert_not_reached();
     } else {
         uint32_t iclass = get_duplex_iclass(encoding);
         unsigned int slot0_subinsn = get_slot0_subinsn(encoding);
@@ -512,6 +511,11 @@ decode_insns(DisasContext *ctx, Insn *insn, uint32_t encoding)
         }
         g_assert_not_reached();
     }
+    /*
+     * invalid/unrecognized opcode; return 1 and let gen_insn() raise an
+     * exception when it sees this empty insn.
+     */
+    return 1;
 }
 
 static void decode_add_endloop_insn(Insn *insn, int loopnum)
-- 
2.34.1


^ permalink raw reply related	[flat|nested] 6+ messages in thread

* RE: [PATCH v3 5/5] target/hexagon: Remove unreachable
  2025-04-07 19:27 ` [PATCH v3 5/5] target/hexagon: Remove unreachable Brian Cain
@ 2025-04-14 17:19   ` ltaylorsimpson
  0 siblings, 0 replies; 6+ messages in thread
From: ltaylorsimpson @ 2025-04-14 17:19 UTC (permalink / raw)
  To: 'Brian Cain', qemu-devel
  Cc: richard.henderson, philmd, matheus.bernardino, ale, anjo,
	marco.liebel, alex.bennee, quic_mburton, sidneym



> -----Original Message-----
> From: Brian Cain <brian.cain@oss.qualcomm.com>
> Sent: Monday, April 7, 2025 1:27 PM
> To: qemu-devel@nongnu.org
> Cc: brian.cain@oss.qualcomm.com; richard.henderson@linaro.org;
> philmd@linaro.org; matheus.bernardino@oss.qualcomm.com; ale@rev.ng;
> anjo@rev.ng; marco.liebel@oss.qualcomm.com; ltaylorsimpson@gmail.com;
> alex.bennee@linaro.org; quic_mburton@quicinc.com;
> sidneym@quicinc.com
> Subject: [PATCH v3 5/5] target/hexagon: Remove unreachable
> 
> We should raise an exception in the event that we encounter a packet that
> can't be correctly decoded, not fault.
> 
> Signed-off-by: Brian Cain <brian.cain@oss.qualcomm.com>
> ---
>  target/hexagon/decode.c | 6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)
> 
> diff --git a/target/hexagon/decode.c b/target/hexagon/decode.c index
> b5ece60450..1db7f1950f 100644
> --- a/target/hexagon/decode.c
> +++ b/target/hexagon/decode.c
> @@ -489,7 +489,6 @@ decode_insns(DisasContext *ctx, Insn *insn, uint32_t
> encoding)
>              insn->iclass = iclass_bits(encoding);
>              return 1;
>          }
> -        g_assert_not_reached();
>      } else {
>          uint32_t iclass = get_duplex_iclass(encoding);
>          unsigned int slot0_subinsn = get_slot0_subinsn(encoding); @@ -512,6
> +511,11 @@ decode_insns(DisasContext *ctx, Insn *insn, uint32_t encoding)
>          }
>          g_assert_not_reached();

Why leave this one rather than raising an exception?

>      }
> +    /*
> +     * invalid/unrecognized opcode; return 1 and let gen_insn() raise an
> +     * exception when it sees this empty insn.
> +     */
> +    return 1;

You should set insn->generate to NULL if you want to guarantee that gen_insn will raise an exception.  A better option is to return a special value that indicates "invalid" and have decode_packet return 0 which will cause decode_and_translate_packet to generate the exception before generating the code for any other instructions in the packet.

Do you have a test case for this?

Taylor




^ permalink raw reply	[flat|nested] 6+ messages in thread

* RE: [PATCH v3 5/5] target/hexagon: Remove unreachable
       [not found] <058801dbad61_68ff5b00_3afe1100_@gmail.com>
@ 2025-04-14 18:09 ` Matheus Tavares Bernardino
  2025-04-14 20:59   ` ltaylorsimpson
  0 siblings, 1 reply; 6+ messages in thread
From: Matheus Tavares Bernardino @ 2025-04-14 18:09 UTC (permalink / raw)
  To: ltaylorsimpson
  Cc: brian.cain, qemu-devel, richard.henderson, philmd,
	matheus.bernardino, ale, anjo, marco.liebel, alex.bennee,
	quic_mburton, sidneym

On Mon, 14 Apr 2025 11:19:38 -0600 <ltaylorsimpson@gmail.com> wrote:
>
> > -----Original Message-----
> > From: Brian Cain <brian.cain@oss.qualcomm.com>
> > Sent: Monday, April 7, 2025 1:27 PM
> > To: qemu-devel@nongnu.org
> > Cc: brian.cain@oss.qualcomm.com; richard.henderson@linaro.org;
> > philmd@linaro.org; matheus.bernardino@oss.qualcomm.com; ale@rev.ng;
> > anjo@rev.ng; marco.liebel@oss.qualcomm.com; ltaylorsimpson@gmail.com;
> > alex.bennee@linaro.org; quic_mburton@quicinc.com;
> > sidneym@quicinc.com
> > Subject: [PATCH v3 5/5] target/hexagon: Remove unreachable
> > 
> > We should raise an exception in the event that we encounter a packet that
> > can't be correctly decoded, not fault.
> > 
> > Signed-off-by: Brian Cain <brian.cain@oss.qualcomm.com>
> > ---
> >  target/hexagon/decode.c | 6 +++++-
> >  1 file changed, 5 insertions(+), 1 deletion(-)
> > 
> > diff --git a/target/hexagon/decode.c b/target/hexagon/decode.c index
> > b5ece60450..1db7f1950f 100644
> > --- a/target/hexagon/decode.c
> > +++ b/target/hexagon/decode.c
> > @@ -489,7 +489,6 @@ decode_insns(DisasContext *ctx, Insn *insn, uint32_t
> > encoding)
> >              insn->iclass = iclass_bits(encoding);
> >              return 1;
> >          }
> > -        g_assert_not_reached();
> >      } else {
> >          uint32_t iclass = get_duplex_iclass(encoding);
> >          unsigned int slot0_subinsn = get_slot0_subinsn(encoding); @@ -512,6
> > +511,11 @@ decode_insns(DisasContext *ctx, Insn *insn, uint32_t encoding)
> >          }
> >          g_assert_not_reached();
> 
> Why leave this one rather than raising an exception?

Good point. I think this one should be removed as well. We have removed it
downstream already.

> >      }
> > +    /*
> > +     * invalid/unrecognized opcode; return 1 and let gen_insn() raise an
> > +     * exception when it sees this empty insn.
> > +     */
> > +    return 1;
> 
> You should set insn->generate to NULL if you want to guarantee that gen_insn
> will raise an exception.

The caller already memset's it to 0 before passing `insn` down.
 
> Do you have a test case for this?

We do have a softmmu test for this downstream. Maybe we can adjust it for
user-mode and upstream it with this patch.


^ permalink raw reply	[flat|nested] 6+ messages in thread

* RE: [PATCH v3 5/5] target/hexagon: Remove unreachable
  2025-04-14 18:09 ` [PATCH v3 5/5] target/hexagon: Remove unreachable Matheus Tavares Bernardino
@ 2025-04-14 20:59   ` ltaylorsimpson
  2025-04-16 22:21     ` Brian Cain
  0 siblings, 1 reply; 6+ messages in thread
From: ltaylorsimpson @ 2025-04-14 20:59 UTC (permalink / raw)
  To: 'Matheus Tavares Bernardino'
  Cc: brian.cain, qemu-devel, richard.henderson, philmd, ale, anjo,
	marco.liebel, alex.bennee, quic_mburton, sidneym



> -----Original Message-----
> From: Matheus Tavares Bernardino
> <matheus.bernardino@oss.qualcomm.com>
> Sent: Monday, April 14, 2025 12:10 PM
> To: ltaylorsimpson@gmail.com
> Cc: brian.cain@oss.qualcomm.com; qemu-devel@nongnu.org;
> richard.henderson@linaro.org; philmd@linaro.org;
> matheus.bernardino@oss.qualcomm.com; ale@rev.ng; anjo@rev.ng;
> marco.liebel@oss.qualcomm.com; alex.bennee@linaro.org;
> quic_mburton@quicinc.com; sidneym@quicinc.com
> Subject: RE: [PATCH v3 5/5] target/hexagon: Remove unreachable
> 
> On Mon, 14 Apr 2025 11:19:38 -0600 <ltaylorsimpson@gmail.com> wrote:
> >
> > > -----Original Message-----
> > > From: Brian Cain <brian.cain@oss.qualcomm.com>
> > > Sent: Monday, April 7, 2025 1:27 PM
> > > To: qemu-devel@nongnu.org
> > > Cc: brian.cain@oss.qualcomm.com; richard.henderson@linaro.org;
> > > philmd@linaro.org; matheus.bernardino@oss.qualcomm.com;
> ale@rev.ng;
> > > anjo@rev.ng; marco.liebel@oss.qualcomm.com;
> > > ltaylorsimpson@gmail.com; alex.bennee@linaro.org;
> > > quic_mburton@quicinc.com; sidneym@quicinc.com
> > > Subject: [PATCH v3 5/5] target/hexagon: Remove unreachable
> > >
> > > We should raise an exception in the event that we encounter a packet
> > > that can't be correctly decoded, not fault.
> > >
> > > Signed-off-by: Brian Cain <brian.cain@oss.qualcomm.com>
> > > ---
> > >  target/hexagon/decode.c | 6 +++++-
> > >  1 file changed, 5 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/target/hexagon/decode.c b/target/hexagon/decode.c index
> > > b5ece60450..1db7f1950f 100644
> > > --- a/target/hexagon/decode.c
> > > +++ b/target/hexagon/decode.c
> > > @@ -489,7 +489,6 @@ decode_insns(DisasContext *ctx, Insn *insn,
> > > uint32_t
> > > encoding)
> > >              insn->iclass = iclass_bits(encoding);
> > >              return 1;
> > >          }
> > > -        g_assert_not_reached();
> > >      } else {
> > >          uint32_t iclass = get_duplex_iclass(encoding);
> > >          unsigned int slot0_subinsn = get_slot0_subinsn(encoding);
> > > @@ -512,6
> > > +511,11 @@ decode_insns(DisasContext *ctx, Insn *insn, uint32_t
> > > +encoding)
> > >          }
> > >          g_assert_not_reached();
> >
> > Why leave this one rather than raising an exception?
> 
> Good point. I think this one should be removed as well. We have removed it
> downstream already.
> 
> > >      }
> > > +    /*
> > > +     * invalid/unrecognized opcode; return 1 and let gen_insn() raise
an
> > > +     * exception when it sees this empty insn.
> > > +     */
> > > +    return 1;
> >
> > You should set insn->generate to NULL if you want to guarantee that
> > gen_insn will raise an exception.
> 
> The caller already memset's it to 0 before passing `insn` down.
> 
> > Do you have a test case for this?
> 
> We do have a softmmu test for this downstream. Maybe we can adjust it for
> user-mode and upstream it with this patch.

Take a look at tests/tcg/hexagon/invalid-slots.c to see how to do this in
linux-user mode.  You'll also need to modify Makefile.target in that
directory.

HTH,
Taylor




^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH v3 5/5] target/hexagon: Remove unreachable
  2025-04-14 20:59   ` ltaylorsimpson
@ 2025-04-16 22:21     ` Brian Cain
  2025-04-16 23:27       ` ltaylorsimpson
  0 siblings, 1 reply; 6+ messages in thread
From: Brian Cain @ 2025-04-16 22:21 UTC (permalink / raw)
  To: ltaylorsimpson, 'Matheus Tavares Bernardino'
  Cc: qemu-devel, richard.henderson, philmd, ale, anjo, marco.liebel,
	alex.bennee, quic_mburton, sidneym


On 4/14/2025 3:59 PM, ltaylorsimpson@gmail.com wrote:
>
>> -----Original Message-----
>> From: Matheus Tavares Bernardino
>> <matheus.bernardino@oss.qualcomm.com>
>> Sent: Monday, April 14, 2025 12:10 PM
>> To: ltaylorsimpson@gmail.com
>> Cc: brian.cain@oss.qualcomm.com; qemu-devel@nongnu.org;
>> richard.henderson@linaro.org; philmd@linaro.org;
>> matheus.bernardino@oss.qualcomm.com; ale@rev.ng; anjo@rev.ng;
>> marco.liebel@oss.qualcomm.com; alex.bennee@linaro.org;
>> quic_mburton@quicinc.com; sidneym@quicinc.com
>> Subject: RE: [PATCH v3 5/5] target/hexagon: Remove unreachable
>>
>> On Mon, 14 Apr 2025 11:19:38 -0600 <ltaylorsimpson@gmail.com> wrote:
>>>> -----Original Message-----
>>>> From: Brian Cain <brian.cain@oss.qualcomm.com>
>>>> Sent: Monday, April 7, 2025 1:27 PM
>>>> To: qemu-devel@nongnu.org
>>>> Cc: brian.cain@oss.qualcomm.com; richard.henderson@linaro.org;
>>>> philmd@linaro.org; matheus.bernardino@oss.qualcomm.com;
>> ale@rev.ng;
>>>> anjo@rev.ng; marco.liebel@oss.qualcomm.com;
>>>> ltaylorsimpson@gmail.com; alex.bennee@linaro.org;
>>>> quic_mburton@quicinc.com; sidneym@quicinc.com
>>>> Subject: [PATCH v3 5/5] target/hexagon: Remove unreachable
>>>>
>>>> We should raise an exception in the event that we encounter a packet
>>>> that can't be correctly decoded, not fault.
>>>>
>>>> Signed-off-by: Brian Cain <brian.cain@oss.qualcomm.com>
>>>> ---
>>>>   target/hexagon/decode.c | 6 +++++-
>>>>   1 file changed, 5 insertions(+), 1 deletion(-)
>>>>
>>>> diff --git a/target/hexagon/decode.c b/target/hexagon/decode.c index
>>>> b5ece60450..1db7f1950f 100644
>>>> --- a/target/hexagon/decode.c
>>>> +++ b/target/hexagon/decode.c
>>>> @@ -489,7 +489,6 @@ decode_insns(DisasContext *ctx, Insn *insn,
>>>> uint32_t
>>>> encoding)
>>>>               insn->iclass = iclass_bits(encoding);
>>>>               return 1;
>>>>           }
>>>> -        g_assert_not_reached();
>>>>       } else {
>>>>           uint32_t iclass = get_duplex_iclass(encoding);
>>>>           unsigned int slot0_subinsn = get_slot0_subinsn(encoding);
>>>> @@ -512,6
>>>> +511,11 @@ decode_insns(DisasContext *ctx, Insn *insn, uint32_t
>>>> +encoding)
>>>>           }
>>>>           g_assert_not_reached();
>>> Why leave this one rather than raising an exception?
>> Good point. I think this one should be removed as well. We have removed it
>> downstream already.


Taylor, is it satisfactory to include that update in a subsequent patch 
series?  Or should this one replace the second unreachable too?


>>
>>>>       }
>>>> +    /*
>>>> +     * invalid/unrecognized opcode; return 1 and let gen_insn() raise
> an
>>>> +     * exception when it sees this empty insn.
>>>> +     */
>>>> +    return 1;
>>> You should set insn->generate to NULL if you want to guarantee that
>>> gen_insn will raise an exception.
>> The caller already memset's it to 0 before passing `insn` down.
>>
>>> Do you have a test case for this?
>> We do have a softmmu test for this downstream. Maybe we can adjust it for
>> user-mode and upstream it with this patch.
> Take a look at tests/tcg/hexagon/invalid-slots.c to see how to do this in
> linux-user mode.  You'll also need to modify Makefile.target in that
> directory.


Matheus provided a linux-user test offline.  I'll include it in an 
updated patch.


>
> HTH,
> Taylor
>
>


^ permalink raw reply	[flat|nested] 6+ messages in thread

* RE: [PATCH v3 5/5] target/hexagon: Remove unreachable
  2025-04-16 22:21     ` Brian Cain
@ 2025-04-16 23:27       ` ltaylorsimpson
  0 siblings, 0 replies; 6+ messages in thread
From: ltaylorsimpson @ 2025-04-16 23:27 UTC (permalink / raw)
  To: 'Brian Cain', 'Matheus Tavares Bernardino'
  Cc: qemu-devel, richard.henderson, philmd, ale, anjo, marco.liebel,
	alex.bennee, quic_mburton, sidneym



> -----Original Message-----
> From: Brian Cain <brian.cain@oss.qualcomm.com>
> Sent: Wednesday, April 16, 2025 4:22 PM
> To: ltaylorsimpson@gmail.com; 'Matheus Tavares Bernardino'
> <matheus.bernardino@oss.qualcomm.com>
> Cc: qemu-devel@nongnu.org; richard.henderson@linaro.org;
> philmd@linaro.org; ale@rev.ng; anjo@rev.ng;
> marco.liebel@oss.qualcomm.com; alex.bennee@linaro.org;
> quic_mburton@quicinc.com; sidneym@quicinc.com
> Subject: Re: [PATCH v3 5/5] target/hexagon: Remove unreachable
> 
> 
> On 4/14/2025 3:59 PM, ltaylorsimpson@gmail.com wrote:
> >
> >> -----Original Message-----
> >> From: Matheus Tavares Bernardino
> >> <matheus.bernardino@oss.qualcomm.com>
> >> Sent: Monday, April 14, 2025 12:10 PM
> >> To: ltaylorsimpson@gmail.com
> >> Cc: brian.cain@oss.qualcomm.com; qemu-devel@nongnu.org;
> >> richard.henderson@linaro.org; philmd@linaro.org;
> >> matheus.bernardino@oss.qualcomm.com; ale@rev.ng; anjo@rev.ng;
> >> marco.liebel@oss.qualcomm.com; alex.bennee@linaro.org;
> >> quic_mburton@quicinc.com; sidneym@quicinc.com
> >> Subject: RE: [PATCH v3 5/5] target/hexagon: Remove unreachable
> >>
> >> On Mon, 14 Apr 2025 11:19:38 -0600 <ltaylorsimpson@gmail.com> wrote:
> >>>> -----Original Message-----
> >>>> From: Brian Cain <brian.cain@oss.qualcomm.com>
> >>>> Sent: Monday, April 7, 2025 1:27 PM
> >>>> To: qemu-devel@nongnu.org
> >>>> Cc: brian.cain@oss.qualcomm.com; richard.henderson@linaro.org;
> >>>> philmd@linaro.org; matheus.bernardino@oss.qualcomm.com;
> >> ale@rev.ng;
> >>>> anjo@rev.ng; marco.liebel@oss.qualcomm.com;
> >>>> ltaylorsimpson@gmail.com; alex.bennee@linaro.org;
> >>>> quic_mburton@quicinc.com; sidneym@quicinc.com
> >>>> Subject: [PATCH v3 5/5] target/hexagon: Remove unreachable
> >>>>
> >>>> We should raise an exception in the event that we encounter a
> >>>> packet that can't be correctly decoded, not fault.
> >>>>
> >>>> Signed-off-by: Brian Cain <brian.cain@oss.qualcomm.com>
> >>>> ---
> >>>>   target/hexagon/decode.c | 6 +++++-
> >>>>   1 file changed, 5 insertions(+), 1 deletion(-)
> >>>>
> >>>> diff --git a/target/hexagon/decode.c b/target/hexagon/decode.c
> >>>> index b5ece60450..1db7f1950f 100644
> >>>> --- a/target/hexagon/decode.c
> >>>> +++ b/target/hexagon/decode.c
> >>>> @@ -489,7 +489,6 @@ decode_insns(DisasContext *ctx, Insn *insn,
> >>>> uint32_t
> >>>> encoding)
> >>>>               insn->iclass = iclass_bits(encoding);
> >>>>               return 1;
> >>>>           }
> >>>> -        g_assert_not_reached();
> >>>>       } else {
> >>>>           uint32_t iclass = get_duplex_iclass(encoding);
> >>>>           unsigned int slot0_subinsn = get_slot0_subinsn(encoding);
> >>>> @@ -512,6
> >>>> +511,11 @@ decode_insns(DisasContext *ctx, Insn *insn, uint32_t
> >>>> +encoding)
> >>>>           }
> >>>>           g_assert_not_reached();
> >>> Why leave this one rather than raising an exception?
> >> Good point. I think this one should be removed as well. We have
> >> removed it downstream already.
> 
> 
> Taylor, is it satisfactory to include that update in a subsequent patch
> series?  Or should this one replace the second unreachable too?

If you just remove that line, it will fall through to the "return 1" below.


> >>
> >>>>       }
> >>>> +    /*
> >>>> +     * invalid/unrecognized opcode; return 1 and let gen_insn() raise
> > an
> >>>> +     * exception when it sees this empty insn.
> >>>> +     */
> >>>> +    return 1;
> >>> You should set insn->generate to NULL if you want to guarantee that
> >>> gen_insn will raise an exception.
> >> The caller already memset's it to 0 before passing `insn` down.
> >>
> >>> Do you have a test case for this?
> >> We do have a softmmu test for this downstream. Maybe we can adjust it
> for
> >> user-mode and upstream it with this patch.
> > Take a look at tests/tcg/hexagon/invalid-slots.c to see how to do this in
> > linux-user mode.  You'll also need to modify Makefile.target in that
> > directory.
> 
> 
> Matheus provided a linux-user test offline.  I'll include it in an
> updated patch.
> 
> 
> >
> > HTH,
> > Taylor
> >
> >



^ permalink raw reply	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2025-04-16 23:29 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
     [not found] <058801dbad61_68ff5b00_3afe1100_@gmail.com>
2025-04-14 18:09 ` [PATCH v3 5/5] target/hexagon: Remove unreachable Matheus Tavares Bernardino
2025-04-14 20:59   ` ltaylorsimpson
2025-04-16 22:21     ` Brian Cain
2025-04-16 23:27       ` ltaylorsimpson
2025-04-07 19:27 [PATCH v3 0/5] misc hexagon patches Brian Cain
2025-04-07 19:27 ` [PATCH v3 5/5] target/hexagon: Remove unreachable Brian Cain
2025-04-14 17:19   ` ltaylorsimpson

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).