| 19 Jun 2026 |
 Atemu | And it's intentional | 15:18:16 |
 magic_rb | :( | 15:18:32 |
 K900 | In reply to @elvishjerricco:matrix.org
isn't the whole point of user namespaces that you get all capabilities, they're just scoped by the ones the creator had in the parent namespace? I don't think so | 15:18:32 |
| K900 | Not when you're doing UID remapping | 15:18:41 |
| K900 | Which we are | 15:18:44 |
| magic_rb | So what is the solution to our nice problem, rtkit? | 15:18:55 |
| Atemu | Long story short: Your options are either to patch your graphics driver to not require the cap for high-prio queues or use Monado | 15:19:11 |
| magic_rb | Cause if im to port jovian to the frame, ill need some other solution | 15:19:30 |
 ElvishJerricco | $ unshare -Uc --keep-caps setpriv -dd | rg -o sys_nice
sys_nice
sys_nice
sys_nice
sys_nice
sys_nice
also works with -r instead of -c
| 15:19:37 |
| magic_rb | Fucking if i have to patch the kernel, so be it | 15:19:42 |
| ElvishJerricco | you definitely have the cap | 15:19:43 |
| K900 | In reply to @magic_rb:matrix.redalder.org
So what is the solution to our nice problem, rtkit? Wait for kernel to get proper DRM scheduling priorities and then rtkit | 15:19:43 |
| ElvishJerricco | I just dunno if that cap is usable for anything meaningful | 15:20:11 |
| K900 | I'm not sure what the Frame will do but I don't think Adreno even has priority contexts | 15:20:15 |
| Atemu | I used to do the former but recently found out that the latter is feasible these days as Monado has a STEAMVR_LH driver | 15:20:34 |
| magic_rb | (I hope youve time k900, ill need your help for the port, i cant do it by myself) | 15:20:52 |
| ElvishJerricco | what do you mean? I think the only way this affects it is that commands like unshare typically drop the caps, but they don't have to (--keep-caps) | 15:21:02 |
| Atemu | Not for high-prio queues. | 15:21:03 |
| magic_rb | STEAMVR_LH | 15:21:17 |
| K900 | In reply to @atemu12:matrix.org
Not for high-prio queues. Hmm, I'm starting to wonder actually | 15:21:19 |
| K900 | This may work | 15:21:26 |
| magic_rb | * | 15:21:27 |
| Atemu | These things are nasty in theory as they can drown out MM | 15:21:37 |
| K900 | Maybe with an upstreamable kernel patch | 15:21:38 |
| Atemu | Nah, upstream doesn't want any of that | 15:21:57 |
| K900 | But also I'm kinda holding out for Valve to figure something out | 15:22:06 |
| Atemu | They think that even SYS_NICE is too lenient | 15:22:11 |
| K900 | Because in a steamrt3 world they will have the same problem | 15:22:17 |
| ElvishJerricco | what kernel patch is needed there? | 15:22:20 |
| magic_rb | Why? | 15:22:32 |
