Re: [PATCH v2 dwarves 4/5] man-pages: document true_signature btf_feature

Yonghong Song <[email protected]>
Newsgroups org.kernel.vger.dwarves,org.kernel.vger.bpf
Message-ID <[email protected]>

On 1/26/26 9:10 AM, Alan Maguire wrote:
> On 26/01/2026 11:21, Matt Bobrowski wrote:
>> On Mon, Jan 26, 2026 at 10:51:31AM +0000, Alan Maguire wrote:
>>> On 26/01/2026 10:02, Matt Bobrowski wrote:
>>>> On Fri, Jan 23, 2026 at 05:26:49PM +0000, Alan Maguire wrote:
>>>>> Ensure non-default "true_signature" feature is documented in
>>>>> the manual page.
>>>>>
>>>>> Signed-off-by: Alan Maguire <[email protected]>
>>>> Acked-by: Matt Bobrowski <[email protected]>
>>>>
>>>>> ---
>>>>>   man-pages/pahole.1 | 5 +++++
>>>>>   1 file changed, 5 insertions(+)
>>>>>
>>>>> diff --git a/man-pages/pahole.1 b/man-pages/pahole.1
>>>>> index 3125de3..90a8f45 100644
>>>>> --- a/man-pages/pahole.1
>>>>> +++ b/man-pages/pahole.1
>>>>> @@ -337,6 +337,11 @@ Supported non-standard features (not enabled for 'default')
>>>>>   	                   of split BTF with a possibly changed base, storing
>>>>>   	                   it in a .BTF.base ELF section.
>>>>>   	global_var         Encode all global variables using BTF_KIND_VAR in BTF.
>>>>> +	true_signature     Encode functions ensuring that binary-level
>>>>> +	                   (rather than source-level) signatures are used;
>>>> 									 ^
>>>> 									 within the generated BTF.
>>>> 									
>>>>> +                           for gcc these are ".isra.0" and ".costprop.0"
>>>>> +                           optimized functions
>>>> For BTF generation, would there ever be a situation whereby we
>>>> wouldn't want true signature support? I'm just trying to understand
>>>> why this isn't defaulted to true.
>>> The key reason is the "." in the name for gcc-optimized functions
>>> "function.isra.0" ; older kernels will reject BTF function names with
>>> a "." in them, so we have to guard against new pahole with the feature
>>> enabled by default being run on older kernels. So by having the newer
>>> kernels request the feature only we avoid this.
>> Oh, yes, that makes sense. Once a new pahole version is released with
>> true_signature support available, will you also be looking to update
>> scripts/Makefile.btf and conditionally enable it based on
>> availability?
> Yep, we're hoping to get some support on the LLVM side for true signatures, but
> absolutely, that's the plan. Because the --btf_features option will simply
> ignore features that it does not know about, this approach works with older
> pahole too without erroring out.

I actually have started to work on this from llvm side. Once Alan's patch
lands (which has true_signature infrastructure implemented), I should be
able to send some patch to address true_signature for llvm-built kernel.
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.