[PATCH v1] PowerPC: Initial SFrame support for 64-bit PowerPC ELF ABI

Abhay Kandpal abhay@linux.ibm.com
Wed Mar 25 09:39:48 GMT 2026 

On 25/03/26 14:17, Jens Remus wrote:
> On 3/24/2026 6:02 PM, Abhay Kandpal wrote:
>> Inspired by s390 commit d27d82f560a8 ("s390: Initial support to generate
>> .sframe from CFI directives in assembler").
>>
>> Support for SFrame on PowerPC 64 is only enabled for the 64-bit PowerPC
>> ELF ABI.
>>
>> The SFrame ABI/arch identifiers SFRAME_ABI_PPC64_ENDIAN_BIG and
>> SFRAME_ABI_PPC64_ENDIAN_LITTLE are introduced for PowerPC 64 big/little
>> endian and added to the SFrame format specification.
>>
>> Port x86-64 commit d7f343eaad3f ("x86-64: Remove sframe relocs against
>> discarded sections").
>>
>> Based on s390 64-bit (s390x) commit 955570f0973d ("s390: Represent FP
>> without RA saved in SFrame").
>>
>> If an architecture uses both SFrame RA and FP tracking SFrame assumes
>> that the RA offset is the 2nd offset and the FP offset is the 3rd offset
>> following a SFrame FRE.  An architecture does not necessarily need to
>> save both on the stack (or in register) at the same time or even at all.
>> SFrame cannot represent FP without RA saved on stack (or in a register),
>> since it cannot distinguish whether the 2nd offset is the RA or FP
>> offset.
>>
>> For PPC64 use an invalid SFrame RA offset from CFA value of zero as
>> padding to represent the FP being saved when the RA is not saved.  This
>> aligns with the existing invalid SFrame fixed RA offset from CFA value
>> of zero.  In a stack tracer this then also naturally falls into place,
>> as it can skip restoring the RA in the topmost frame, if both the fixed
>> RA offset (from SFrame header) and the RA offset (from FDE) are zero,
>> without any need to test architecture-specific flags.
>>
>> include/
>> 	* sframe.h (SFRAME_ABI_PPC64_ENDIAN_BIG,
>> 	SFRAME_ABI_PPC64_ENDIAN_LITTLE): Define.
>> 	* sframe-api.h (sframe_fre_get_ra_offset): Add PPC64 to comment
>> 	for RA offset value of SFRAME_FRE_RA_OFFSET_INVALID indicating
>> 	that the RA is not saved.
>>
>> libsframe/
>> 	* sframe.c (need_swapping): Handle SFRAME_ABI_PPC64_ENDIAN_BIG
>> 	and SFRAME_ABI_PPC64_ENDIAN_LITTLE.
>> 	* doc/sframe-spec.texi (SFRAME_ABI_PPC64_ENDIAN_BIG,
>> 	SFRAME_ABI_PPC64_ENDIAN_LITTLE, PPC64): Document SFrame ABI/arch
>> 	identifier for PPC64, add references to 64-bit PowerPC
>> 	architecture, and document PPC64-specifics.
>> 	* sframe-dump.c (ppc64_sframe_abi_reg_map): Define register map
>> 	for PPC64.
>> 	(sframe_get_reg_name): Use PPC64 register map for PPC64.
>> 	(dump_sframe_func_fres_simple): Update comment to mention PPC64
>> 	for RA offset padding.
>>
>> gas/
>> 	* NEWS: Mention powerpc64 support to generate SFrame from CFI
>> 	directives.
>> 	* config/tc-ppc.h (support_sframe_p, SFRAME_CFA_SP_REG,
>> 	SFRAME_CFA_FP_REG, SFRAME_CFA_RA_REG, sframe_ra_tracking_p,
>> 	sframe_cfa_ra_offset, sframe_get_abi_arch,
>> 	sframe_support_flex_fde_p): Define.
>> 	* config/tc-ppc.c (ppc_support_sframe_p,
>> 	ppc_sframe_get_abi_arch): New functions.  Return whether SFrame
>> 	is supported and the SFrame ABI/arch identifier.
>> 	gen-sframe.c (get_fre_dataword_count): For PPC64 account padding
>> 	RA offset, if FP without RA saved.
>> 	(sframe_get_fre_dataword_size): Likewise.
>> 	(output_sframe_row_entry_datawords): For PPC64 write a padding
>> 	RA offset, if FP without RA needs to be represented.
>> 	(sframe_do_fde): Enable FP without RA saved to be represented
>> 	on PPC64.
>>
>> gas/testsuite/
>> 	* gas/cfi-sframe/cfi-sframe.exp: Enable common SFrame tests
>> 	on PPC64.
>> 	* gas/cfi-sframe/cfi-sframe-common-8.s: Use registers 2 and 3,
>> 	as register 1 is SP on PPC64.
>> 	* gas/cfi-sframe/cfi-sframe-common-10.s: Likewise.
>> 	* gas/cfi-sframe/cfi-sframe-ppc64-1.s: New test.
>> 	* gas/cfi-sframe/cfi-sframe-ppc64-1.d: Likewise.
>> 	* gas/cfi-sframe/cfi-sframe-ppc64-3.s: New test.
>> 	* gas/cfi-sframe/cfi-sframe-ppc64-3.d: Likewise.
>> 	* gas/cfi-sframe/cfi-sframe-ppc64-err-3.s: New test.
>> 	* gas/cfi-sframe/cfi-sframe-ppc64-err-3.d: Likewise.
>   	* gas/cfi-sframe/cfi-sframe-ppc64-fpra-offset-1.s: New test.
>   	* gas/cfi-sframe/cfi-sframe-ppc64-fpra-offset-1.d: Likewise.
>
>> 	* gas/cfi-sframe/cfi-sframe-ppc64-fpra-offset-2.s: New test.
>> 	* gas/cfi-sframe/cfi-sframe-ppc64-fpra-offset-2.d: Likewise.
>> 	* gas/cfi-sframe/cfi-sframe-ppc64-fpra-register-2.s: New test.
>> 	* gas/cfi-sframe/cfi-sframe-ppc64-fpra-register-2.d: Likewise.
>> 	* gas/cfi-sframe/cfi-sframe-ppc64-non-spfp-cfa-2.s: New test.
>> 	* gas/cfi-sframe/cfi-sframe-ppc64-non-spfp-cfa-2.d: Likewise.
>> 	* gas/cfi-sframe/cfi-sframe-ppc64-pr33756.s: New test.
>> 	* gas/cfi-sframe/cfi-sframe-ppc64-pr33756.d: Likewise.
>> 	* gas/cfi-sframe/cfi-sframe-ppc64-ra-undefined-1.s: New test.
>> 	* gas/cfi-sframe/cfi-sframe-ppc64-ra-undefined-1.d: Likewise.
>>
>> bfd/
>> 	* elf64-ppc.c (ppc64_elf_relocate_section): Remove sframe relocs
>> 	against discarded sections.
>>
>> Co-authored-by: Jens Remus<jremus@linux.ibm.com>
>> Signed-off-by: Abhay Kandpal<abhay@linux.ibm.com>
> LGTM.  Please see my minor comments.
>
> Let's see what Indu has to say. :-)
>
>> diff --git a/gas/NEWS b/gas/NEWS
>> index e384d1135c0..fbc755861ce 100644
>> --- a/gas/NEWS
>> +++ b/gas/NEWS
>> @@ -1,5 +1,10 @@
>>   -*- text -*-
>>   
>> +Changes in 2.47:
>> +
> Nit: IIUC above line is added by the binutils maintainer before a new
> release, so please drop that.

sure

>
>> +* Add support to generate SFrame stack trace information (.sframe)
>> +  from CFI directives on PowerPC 64-bit (powerpc64).
>> +
>>   Changes in 2.46:
>>   
>>   * Add support for AMD Zen6 processor.
>> diff --git a/gas/testsuite/gas/cfi-sframe/cfi-sframe-ppc64-fpra-offset-1.s b/gas/testsuite/gas/cfi-sframe/cfi-sframe-ppc64-fpra-offset-1.s
>> new file mode 100644
>> index 00000000000..6053b54534a
>> --- /dev/null
>> +++ b/gas/testsuite/gas/cfi-sframe/cfi-sframe-ppc64-fpra-offset-1.s
>> @@ -0,0 +1,17 @@
>> +	.cfi_sections .sframe
>> +	.cfi_startproc
>> +	mflr 0
>> +	std 0,112(1)
>> +	.cfi_offset 65, 112
>> +	std 31,88(1)
>> +	.cfi_offset 31, 88
> Is it intentional that RA (= LR) and FP are saved at non-default
> offsets?  Are these within the red zone on PPC64?

The test intention is to verify 'U' padding mechanism.
Although the offsets 112, 88 are not in red zone, they are above SP in the caller's frame area, I will modify
these offset for the both the test cases to use standard PPC64 ABI offset to make test more realistic and consistent across the PPC test cases.

>
>> +	li 31,0
>> +	li 0,0
>> +.Lreturn:
>> +	ld 31,88(1)
>> +	.cfi_restore 31
>> +	ld 0,112(1)
>> +	mtlr 0
>> +	.cfi_restore 65
>> +	blr
>> +	.cfi_endproc
>
>> diff --git a/gas/testsuite/gas/cfi-sframe/cfi-sframe-ppc64-fpra-offset-2.s b/gas/testsuite/gas/cfi-sframe/cfi-sframe-ppc64-fpra-offset-2.s
>> new file mode 100644
>> index 00000000000..67798c24191
>> --- /dev/null
>> +++ b/gas/testsuite/gas/cfi-sframe/cfi-sframe-ppc64-fpra-offset-2.s
>> @@ -0,0 +1,18 @@
>> +	.cfi_sections .sframe
>> +	.cfi_startproc
>> +	# No stack allocation, CFA = SP+0 implicitly
>> +	std 31,88(1)            # Save FP at SP+88
>> +	.cfi_offset 31, 88      # FP at CFA+88
>> +	mflr 0                  # Get LR
>> +	std 0,112(1)            # Save RA at SP+112
>> +	.cfi_offset 65, 112     # RA at CFA+112
> Likewise (non-default offsets intentional?).
>
>> +	li 31,0                 # Clear registers
>> +	li 0,0
>> +.Lreturn:
>> +	ld 0,112(1)             # Restore RA
>> +	mtlr 0
>> +	.cfi_restore 65
>> +	ld 31,88(1)             # Restore FP
>> +	.cfi_restore 31
>> +	blr
>> +	.cfi_endproc
> Thanks and regards,
> Jens
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://sourceware.org/pipermail/binutils/attachments/20260325/0940d9e0/attachment.htm>

More information about the Binutils mailing list