Re: [PING][PATCH,v5][AArch64] MTE corefile support

[Luis Machado <luis.machado@linaro.org>]

On 8/1/21 12:44 PM, Joel Brobecker wrote:
>> On 2021-07-29 2:10 p.m., Catalin Marinas wrote:
>>> I'm more worried about GDB 11 reading the MTE notes incorrectly rather
>>> than not reading them at all.
>>
>> Yeah, if a GDB 11 that knows how to read the pre-standardized notes does
>> not fail gracefully when encountering the standardized notes, it would
>> be bad.  If it just skipped them and said "I don't recognize them", it
>> would be ok.
> 
> Agreed.
> 
> Is there a way we could check the notes? I looked at the patch,
> and it doesn't seem like there is any kind of versioning or
> identifier we could use to determine which version of the notes
> we get. But perhaps we can at least do some kind of consistency

We have a generic header containing NT_MEMTAG and an arch-specific 
header identifier containing NT_MEMTAG_TYPE_AARCH_MTE. The arch-specific 
identifier can be versioned as needed. We could, for example, have 
NT_MEMTAG_TYPE_AARCH_MTE_V2.

> check, such as for instance comparing the size of the section
> with the contents of the section header, and generate an error
> if that doesn't match. The error might not be as informative,
> but perhaps still good enough?

On further discussion with the Linux Kernel developers, we're going to 
take the safe path and will wait for this part to be under review on the 
Linux Kernel ML before adopting it in GDB.

> 
>> Oh, ok, I thought it had already gone through some review cycles.  Then
>> I'll leave it to you all to decide, those who have some stakes in this.
>