* [PATCH] xtensa: initialize call_abi in xtensa_tdep
@ 2015-06-06 22:06 Max Filippov
2015-08-20 7:38 ` Max Filippov
2015-08-20 12:09 ` Joel Brobecker
0 siblings, 2 replies; 6+ messages in thread
From: Max Filippov @ 2015-06-06 22:06 UTC (permalink / raw)
To: gdb-patches; +Cc: Maxim Grigoriev, Woody LaRue, Marc Gauthier, Max Filippov
Use XSHAL_ABI value provided by xtensa-config.h to correctly initialize
xtensa_tdep.call_abi
This fixes calls to functions from GDB that otherwise fail with the
following assertion in call0 configuration:
gdb/regcache.c:602: internal-error: regcache_raw_read: Assertion
`regnum >= 0 && regnum < regcache->descr->nr_raw_registers' failed.
gdb/
* xtensa-tdep.h (XTENSA_GDBARCH_TDEP_INSTANTIATE): Initialize
call_abi using XSHAL_ABI macro.
---
gdb/xtensa-tdep.h | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/gdb/xtensa-tdep.h b/gdb/xtensa-tdep.h
index adacaf8..3b6ea66 100644
--- a/gdb/xtensa-tdep.h
+++ b/gdb/xtensa-tdep.h
@@ -246,7 +246,8 @@ struct gdbarch_tdep
.spill_location = -1, \
.spill_size = (spillsz), \
.unused = 0, \
- .call_abi = 0, \
+ .call_abi = (XSHAL_ABI == XTHAL_ABI_CALL0) ? \
+ CallAbiCall0Only : CallAbiDefault, \
.debug_interrupt_level = XCHAL_DEBUGLEVEL, \
.icache_line_bytes = XCHAL_ICACHE_LINESIZE, \
.dcache_line_bytes = XCHAL_DCACHE_LINESIZE, \
--
1.8.1.4
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] xtensa: initialize call_abi in xtensa_tdep
2015-06-06 22:06 [PATCH] xtensa: initialize call_abi in xtensa_tdep Max Filippov
@ 2015-08-20 7:38 ` Max Filippov
2015-08-20 12:09 ` Joel Brobecker
1 sibling, 0 replies; 6+ messages in thread
From: Max Filippov @ 2015-08-20 7:38 UTC (permalink / raw)
To: gdb-patches; +Cc: Maxim Grigoriev, Woody LaRue, Marc Gauthier, Max Filippov
On Sun, Jun 7, 2015 at 1:05 AM, Max Filippov <jcmvbkbc@gmail.com> wrote:
> Use XSHAL_ABI value provided by xtensa-config.h to correctly initialize
> xtensa_tdep.call_abi
> This fixes calls to functions from GDB that otherwise fail with the
> following assertion in call0 configuration:
>
> gdb/regcache.c:602: internal-error: regcache_raw_read: Assertion
> `regnum >= 0 && regnum < regcache->descr->nr_raw_registers' failed.
>
> gdb/
> * xtensa-tdep.h (XTENSA_GDBARCH_TDEP_INSTANTIATE): Initialize
> call_abi using XSHAL_ABI macro.
> ---
> gdb/xtensa-tdep.h | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/gdb/xtensa-tdep.h b/gdb/xtensa-tdep.h
> index adacaf8..3b6ea66 100644
> --- a/gdb/xtensa-tdep.h
> +++ b/gdb/xtensa-tdep.h
> @@ -246,7 +246,8 @@ struct gdbarch_tdep
> .spill_location = -1, \
> .spill_size = (spillsz), \
> .unused = 0, \
> - .call_abi = 0, \
> + .call_abi = (XSHAL_ABI == XTHAL_ABI_CALL0) ? \
> + CallAbiCall0Only : CallAbiDefault, \
> .debug_interrupt_level = XCHAL_DEBUGLEVEL, \
> .icache_line_bytes = XCHAL_ICACHE_LINESIZE, \
> .dcache_line_bytes = XCHAL_DCACHE_LINESIZE, \
> --
Ping?
--
Thanks.
-- Max
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] xtensa: initialize call_abi in xtensa_tdep
2015-06-06 22:06 [PATCH] xtensa: initialize call_abi in xtensa_tdep Max Filippov
2015-08-20 7:38 ` Max Filippov
@ 2015-08-20 12:09 ` Joel Brobecker
2015-08-20 12:47 ` Max Filippov
1 sibling, 1 reply; 6+ messages in thread
From: Joel Brobecker @ 2015-08-20 12:09 UTC (permalink / raw)
To: Max Filippov; +Cc: gdb-patches, Maxim Grigoriev, Woody LaRue, Marc Gauthier
On Sun, Jun 07, 2015 at 01:05:36AM +0300, Max Filippov wrote:
> Use XSHAL_ABI value provided by xtensa-config.h to correctly initialize
> xtensa_tdep.call_abi
> This fixes calls to functions from GDB that otherwise fail with the
> following assertion in call0 configuration:
>
> gdb/regcache.c:602: internal-error: regcache_raw_read: Assertion
> `regnum >= 0 && regnum < regcache->descr->nr_raw_registers' failed.
>
> gdb/
> * xtensa-tdep.h (XTENSA_GDBARCH_TDEP_INSTANTIATE): Initialize
> call_abi using XSHAL_ABI macro.
I am not sure I understand the patch.
The first thing I should mention is that, unless what your patch
suggests, the code as I see it on master currently initializes call_abi
to CallAbiDefault:
.call_abi = CallAbiDefault, \
And looking at the definitions of XSHAL_ABI and XTHAL_ABI_CALL0
in include/xtensa-config.h, those definitions seem to be entirely
static...
#undef XSHAL_ABI
#undef XTHAL_ABI_WINDOWED
#undef XTHAL_ABI_CALL0
#define XSHAL_ABI XTHAL_ABI_WINDOWED
#define XTHAL_ABI_WINDOWED 0
#define XTHAL_ABI_CALL0 1
... and resolving to distinct values. So the condition would always
be false, resulting in call_abi always being set to CallAbiDefault.
Am I missing something?
> gdb/xtensa-tdep.h | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/gdb/xtensa-tdep.h b/gdb/xtensa-tdep.h
> index adacaf8..3b6ea66 100644
> --- a/gdb/xtensa-tdep.h
> +++ b/gdb/xtensa-tdep.h
> @@ -246,7 +246,8 @@ struct gdbarch_tdep
> .spill_location = -1, \
> .spill_size = (spillsz), \
> .unused = 0, \
> - .call_abi = 0, \
> + .call_abi = (XSHAL_ABI == XTHAL_ABI_CALL0) ? \
> + CallAbiCall0Only : CallAbiDefault, \
Small style issue: binary operators should be at the start of
the next line, rather than the end of the current line.
Also, for multi-line conditions like that, the GNU Coding Standard
asks that we help automatic code formatters by wrapping the expression
inside parentheses. These are redundant for the compiler, but help
code formatters. Since the parens around the == operator are unecessary,
I would have written the above:
.call_abi = (XSHAL_ABI == XTHAL_ABI_CALL0 \
? CallAbiCall0Only : CallAbiDefault), \
People sometime even write it as the following, finding it clearer
(although a matter of taste, of course):
.call_abi = (XSHAL_ABI == XTHAL_ABI_CALL0 \
? CallAbiCall0Only \
: CallAbiDefault), \
I personally have no preference.
> .debug_interrupt_level = XCHAL_DEBUGLEVEL, \
> .icache_line_bytes = XCHAL_ICACHE_LINESIZE, \
> .dcache_line_bytes = XCHAL_DCACHE_LINESIZE, \
> --
> 1.8.1.4
--
Joel
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] xtensa: initialize call_abi in xtensa_tdep
2015-08-20 12:09 ` Joel Brobecker
@ 2015-08-20 12:47 ` Max Filippov
2015-08-20 12:53 ` Joel Brobecker
0 siblings, 1 reply; 6+ messages in thread
From: Max Filippov @ 2015-08-20 12:47 UTC (permalink / raw)
To: Joel Brobecker; +Cc: gdb-patches, Maxim Grigoriev, Woody LaRue, Marc Gauthier
On Thu, Aug 20, 2015 at 3:09 PM, Joel Brobecker <brobecker@adacore.com> wrote:
> On Sun, Jun 07, 2015 at 01:05:36AM +0300, Max Filippov wrote:
>> Use XSHAL_ABI value provided by xtensa-config.h to correctly initialize
>> xtensa_tdep.call_abi
>> This fixes calls to functions from GDB that otherwise fail with the
>> following assertion in call0 configuration:
>>
>> gdb/regcache.c:602: internal-error: regcache_raw_read: Assertion
>> `regnum >= 0 && regnum < regcache->descr->nr_raw_registers' failed.
>>
>> gdb/
>> * xtensa-tdep.h (XTENSA_GDBARCH_TDEP_INSTANTIATE): Initialize
>> call_abi using XSHAL_ABI macro.
>
> I am not sure I understand the patch.
>
> The first thing I should mention is that, unless what your patch
> suggests, the code as I see it on master currently initializes call_abi
> to CallAbiDefault:
>
> .call_abi = CallAbiDefault, \
>
> And looking at the definitions of XSHAL_ABI and XTHAL_ABI_CALL0
> in include/xtensa-config.h, those definitions seem to be entirely
> static...
>
> #undef XSHAL_ABI
> #undef XTHAL_ABI_WINDOWED
> #undef XTHAL_ABI_CALL0
> #define XSHAL_ABI XTHAL_ABI_WINDOWED
> #define XTHAL_ABI_WINDOWED 0
> #define XTHAL_ABI_CALL0 1
>
> ... and resolving to distinct values. So the condition would always
> be false, resulting in call_abi always being set to CallAbiDefault.
>
> Am I missing something?
The file include/xtensa-config.h is meant to be replaced by processor-specific
version where XSHAL_ABI definition may be different. That's why I mention
'call0 configuration' in the description: when xtensa core is configured without
windowed registers it will necessarily have call0 ABI.
>> gdb/xtensa-tdep.h | 3 ++-
>> 1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/gdb/xtensa-tdep.h b/gdb/xtensa-tdep.h
>> index adacaf8..3b6ea66 100644
>> --- a/gdb/xtensa-tdep.h
>> +++ b/gdb/xtensa-tdep.h
>> @@ -246,7 +246,8 @@ struct gdbarch_tdep
>> .spill_location = -1, \
>> .spill_size = (spillsz), \
>> .unused = 0, \
>> - .call_abi = 0, \
>> + .call_abi = (XSHAL_ABI == XTHAL_ABI_CALL0) ? \
>> + CallAbiCall0Only : CallAbiDefault, \
>
> Small style issue: binary operators should be at the start of
> the next line, rather than the end of the current line.
I'll send fixed v2.
--
Thanks.
-- Max
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] xtensa: initialize call_abi in xtensa_tdep
2015-08-20 12:47 ` Max Filippov
@ 2015-08-20 12:53 ` Joel Brobecker
2015-08-20 13:00 ` Max Filippov
0 siblings, 1 reply; 6+ messages in thread
From: Joel Brobecker @ 2015-08-20 12:53 UTC (permalink / raw)
To: Max Filippov; +Cc: gdb-patches, Maxim Grigoriev, Woody LaRue, Marc Gauthier
> The file include/xtensa-config.h is meant to be replaced by
> processor-specific version where XSHAL_ABI definition may be
> different. That's why I mention 'call0 configuration' in the
> description: when xtensa core is configured without
> windowed registers it will necessarily have call0 ABI.
This is the type of property that is normally dynamically determined.
Is there a way to determine that info from the executable? That would
be the proper way of fixing this issue.
--
Joel
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: [PATCH] xtensa: initialize call_abi in xtensa_tdep
2015-08-20 12:53 ` Joel Brobecker
@ 2015-08-20 13:00 ` Max Filippov
0 siblings, 0 replies; 6+ messages in thread
From: Max Filippov @ 2015-08-20 13:00 UTC (permalink / raw)
To: Joel Brobecker; +Cc: gdb-patches, Maxim Grigoriev, Woody LaRue, Marc Gauthier
On Thu, Aug 20, 2015 at 3:53 PM, Joel Brobecker <brobecker@adacore.com> wrote:
>> The file include/xtensa-config.h is meant to be replaced by
>> processor-specific version where XSHAL_ABI definition may be
>> different. That's why I mention 'call0 configuration' in the
>> description: when xtensa core is configured without
>> windowed registers it will necessarily have call0 ABI.
>
> This is the type of property that is normally dynamically determined.
> Is there a way to determine that info from the executable? That would
> be the proper way of fixing this issue.
I agree with that, but currently we can't distinguish executables with different
call ABI.
--
Thanks.
-- Max
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2015-08-20 13:00 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2015-06-06 22:06 [PATCH] xtensa: initialize call_abi in xtensa_tdep Max Filippov
2015-08-20 7:38 ` Max Filippov
2015-08-20 12:09 ` Joel Brobecker
2015-08-20 12:47 ` Max Filippov
2015-08-20 12:53 ` Joel Brobecker
2015-08-20 13:00 ` Max Filippov
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox