Re: [PATCH v2 7/7] aarch64,arm: Move branch-protection data to targets

["Richard Earnshaw (lists)" <Richard.Earnshaw@arm.com>]

On 11/01/2024 14:43, Szabolcs Nagy wrote:
> The 12/07/2023 13:13, Richard Earnshaw wrote:
>> On 03/11/2023 15:36, Szabolcs Nagy wrote:
>>> 	* config/aarch64/aarch64.cc (aarch_handle_no_branch_protection): Copy.
>>> 	(aarch_handle_standard_branch_protection): Copy.
>>> 	(aarch_handle_pac_ret_protection): Copy.
>>> 	(aarch_handle_pac_ret_leaf): Copy.
>>> 	(aarch_handle_pac_ret_b_key): Copy.
>>> 	(aarch_handle_bti_protection): Copy.
>>
>> I think all of the above functions that have been moved back from
>> aarch-common should be renamed back to aarch64_..., unless they are directly
>> referenced statically by code in aarch-common.c.
> 
> done.
> 
>>> +const struct aarch_branch_protect_type aarch_branch_protect_types[] = {
>>
>> can this be made static now?  And maybe pass the structure as a parameter if
>> that's not done already.
> 
> done in v4.
> 
>> It would be nice if, when we raise an error, we could print out the list of
>> valid options (and modifiers), much like we do on Arm for -march/-mcpu.
>>
>> eg.
>> $ gcc -mcpu=crotex-a8
>> cc1: error: unrecognised -mcpu target: crotex-a8
>> cc1: note: valid arguments are: arm8 arm810 strongarm strongarm110 fa526
>> [...rest of list]; did you mean ‘cortex-a8’?
> 
> i implemented this with candidates_list_and_hint but it does
> not work very well if the typo is in a subtype, so i think
> this should be done in a separate patch if at all.
> 

I'd build the candidates list from all the types + subtypes, so that the suggestion code has a full list to pick from; but fair enough.

R.