[RFC,V2 0/5] Add support for MTE stack tagging

[Jan Beulich]

On 11.04.2025 11:40, Richard Sandiford wrote:
> Jan Beulich <jbeulich@suse.com> writes:
>> On 11.04.2025 06:16, Indu Bhagat wrote:
>>> Hello,
>>>
>>> Not much has moved from the version of the patches first posted earlier
>>> in December 2024 (link
>>> https://inbox.sourceware.org/binutils/20241205225837.2912674-1-indu.bhagat@oracle.com/).
>>> So this is mostly a rebased version of V1 with some nits fixed.
>>>
>>> The commit log of each commit has some of unresolved questions listed.
>>> Most of the questions are around the implementation of the two dynamic
>>> tags (related to MTE stack tagging).  The patchset addresses
>>> three different aspects (i.e., .cfi_mte_tagged_frame CFI directive,
>>> DT_AARCH64_MEMTAG_MODE and DT_AARCH64_MEMTAG_STACK) related to MTE stack
>>> tagging.
>>>
>>> Thanks,
>>>
>>> Indu Bhagat (5):
>>>   binutils: make read_cie aware of new augmentation char 'G'
>>
>> So where's this 'G' documented? Something must have been added somewhere,
>> if so far the character wasn't recognized?
> 
> It's documented in:
> 
>   https://github.com/ARM-software/abi-aa/blob/main/aadwarf64/aadwarf64.rst

Ah, and that's also where 'B' apparently comes from. Can we really assume
arch extension chars to have uniform behavior everywhere? I.e. for the
purpose of read_cie(), no "operands"?

Jan