Brutkey

Petr Tesařík
@ptesarik@infosec.exchange

If you feel insufficiently frustrated developing stuff, try root-causing performance regressions. Plenty of frustration for everyone!

For example, I am now chasing a netperf regression between 6.4 and 6.12 where perf suggests that raw_copy_from_user() is now slower for some reason. Lacking any better idea, I wrote a small misc driver to stress raw_copy_from_user(). Of course, this raw_copy_from_user() runs equally fast in both kernels. Only the other raw_copy_from_user() which is embedded in copy_from_user_iter() is slower in the newer kernel.

Vlastimil Babka 🇨🇿🇪🇺🇺🇦
@vbabka@mastodon.social

@ptesarik@infosec.exchange I wrote a kunit "test" to stress slab allocations/frees and it shows my changes are making them faster. I'll pretend I never saw your toot.

Petr Tesařík
@ptesarik@infosec.exchange

@vbabka@mastodon.social I'll pretend I never saw your kunit tests.

Lorenzo
@ljs@mastodonapp.uk

@ptesarik@infosec.exchange @vbabka@mastodon.social I'll pretend I never saw your bare feet

Petr Tesařík
@ptesarik@infosec.exchange

@ljs@mastodonapp.uk @vbabka@mastodon.social This is very generous of you. Unfortunately, I am unable to return the favour and pretend I never saw this:

Lorenzo
@ljs@mastodonapp.uk

@ptesarik@infosec.exchange @vbabka@mastodon.social oh you're gonna see a lot more of that bro

Lorenzo
@ljs@mastodonapp.uk

@ptesarik@infosec.exchange @vbabka@mastodon.social

Lorenzo
@ljs@mastodonapp.uk

@ptesarik@infosec.exchange @vbabka@mastodon.social turns out I can't train my ugly face in the gym 🤣

Petr Tesařík
@ptesarik@infosec.exchange

@ljs@mastodonapp.uk @vbabka@mastodon.social Nooooo! I used to sleep so well…

Lorenzo
@ljs@mastodonapp.uk

@ptesarik@infosec.exchange @vbabka@mastodon.social does it help you sleep to know I work in a core kernel area too?

Good luck with that system stability bro!

Petr Tesařík
@ptesarik@infosec.exchange

@ljs@mastodonapp.uk @vbabka@mastodon.social Bro, a long way to go before you can match my “cleanup” in SWIOTLB which broke pretty much all DMA transfers going through an IOMMU. It took two releases and the effort of three other guys to disentangle and fix the mess.

EDIT: But it's worth it—I am now officially an expert.