Slow (higher scheduling latency?) system after S3 - regression 4.17 -> 4.19 ?

["Marek Marczykowski-Górecki" <marmarek@invisiblethingslab.com>]

[-- Attachment #1: Type: text/plain, Size: 1328 bytes --]

Hi,

After updating from Xen 4.17 to 4.19 it appears that I have some
performance issues if system was suspended before. For now this is just
vague observation, as I don't have much data yet.

Generally, the user observable effect is that audio playback (over
Qubes's PV audio thing) becomes very choppy, beyond usefulness (cannot
make audio calls anymore, music/video playback also becomes useless with
some players. For me this suggests some increased latency in scheduling
- maybe some timer fires too late?

This doesn't happen always, but I'd say quite often after S3, sometimes
may need two or three suspend cycles to trigger the issue. But it's
possible it is also some other factor, like which pCPU the VM gets
scheduled on, maybe? But for now I didn't hit this issue before
performing any S3. I don't have a reliable reproducer yet.

While I write it's related to 4.17 -> 4.19 update, there is also slight
possibility it's related to hardware change (CFL -> MTL). But, it's
unlikely, since I'd probably get much more bug reports about that (the
stable Qubes version uses 4.17, and we have many users on MTL).

What would be the method to collect some useful debugging info? Some
debug-keys (r? a?), some cmdline options?

-- 
Best Regards,
Marek Marczykowski-Górecki
Invisible Things Lab

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]