On May 27, 2021 6:08:33 PM GMT-03:00, Jiri Olsa [off-list ref] wrote:

On Thu, May 27, 2021 at 01:41:13PM -0700, Andrii Nakryiko wrote:

quoted

On Thu, May 27, 2021 at 12:57 PM Arnaldo [off-list ref]

wrote:

quoted

quoted

On May 27, 2021 4:14:17 PM GMT-03:00, Andrii Nakryiko

[off-list ref] wrote:

quoted

quoted

quoted

On Thu, May 27, 2021 at 12:06 PM Arnaldo [off-list ref]
wrote:

quoted

On May 27, 2021 1:54:40 PM GMT-03:00, Andrii Nakryiko

[off-list ref] wrote:

quoted

quoted

On Thu, May 27, 2021 at 8:20 AM Arnaldo Carvalho de Melo
[off-list ref] wrote:

quoted

Hi guys,

        Its important to have 1.22 out of the door ASAP, so

please

quoted

quoted

quoted

quoted

quoted

clone

quoted

what is in tmp.master and report your results.

Hey Arnaldo,

If we are going to make pahole 1.22 a new mandatory minimal

version

quoted

quoted

quoted

of

quoted

quoted

pahole, I think we should take a little bit of time and fix

another

quoted

quoted

quoted

quoted

quoted

problematic issue and clean up Kbuild significantly.

We discussed this before, it would be great to have an ability

to

quoted

quoted

quoted

dump

quoted

quoted

generated BTF into a separate file instead of modifying vmlinux

image

quoted

quoted

in place. I'd say let's try to push for [0] to land as a

temporary

quoted

quoted

quoted

quoted

quoted

work around to buy us a bit of time to implement this feature.

Then,

quoted

quoted

quoted

quoted

quoted

when pahole 1.22 is released and packaged into major distros,

we can

quoted

quoted

quoted

quoted

quoted

follow up in kernel with Kbuild clean ups and making pahole

1.22

quoted

quoted

quoted

quoted

quoted

mandatory.

What do you think? If anyone agrees, please consider chiming in

on

quoted

quoted

quoted

the

quoted

quoted

above thread ([0]).

There's multiple fixes that affects lots of stakeholders, so I'm

more

quoted

quoted

quoted

inclined to release 1.22 sooner rather than later.

quoted

If anyone has cycles right now to work on that detached BTF

feature,

quoted

quoted

quoted

releasing 1.23 as soon as that feature is complete and tested

shouldn't

quoted

quoted

quoted

be a problem.

quoted

Then 1.23 the mandatory minimal version.

Wdyt?

If we make 1.22 mandatory there will be no good reason to make

1.23

quoted

quoted

quoted

mandatory again. So I will have absolutely no inclination to work

on

quoted

quoted

quoted

this, for example. So we are just wasting a chance to clean up the
Kbuild story w.r.t. pahole. And we are talking about just a few

days

quoted

quoted

quoted

at most, while we do have a reasonable work around on the kernel

side.

quoted

quoted

So there were patches for stop using objcopy, which we thought

could uncover some can of worms, were there patches for the detached
BTF  file?

quoted

No, there weren't, if I remember correctly. What's the concern,
though? That detached BTF file isn't even an ELF, so it's
btf__get_raw_data() and write it to the file. Done.

heya,
I probably overlooked this, but are there more details about that
detached BTF file feature somewhere? 

Look in the dwarves mailing list archives at lore, but it's just a new option to ask for the BTF data to be written to a file instead of to an ELF section, that will simplify the series of steps in the kernel building process.

I'll cook a patch early tomorrow.

- Arnaldo

-- 
Sent from my Android device with K-9 Mail. Please excuse my brevity.