| 19 Jun 2026 |
 magic_rb | (Signal and irc are not in the same space whatsoever) | 15:38:21 |
| magic_rb | (Xmpp is even more niche than this) | 15:38:28 |
 Atemu | I heard the ATproto folks are working on something that sounded promising | 15:38:38 |
| Atemu | And Fluxer had ambitions for federation | 15:39:02 |
| magic_rb | I mean, i dont. Matrix is great. Technically too. Its just extremely opaque. I wish my client would be able to tell me the delivery times to different HSs over federation. If it had better feedback it wouldnt be nearly so baf | 15:39:16 |
| magic_rb | The issue isnt that messages take a second to deliver, or that federation can fail. The issue is that unless youre staring at your server logs, you wont know | 15:39:50 |
| Atemu | I think we're getting OT now | 15:39:54 |
| magic_rb | Anyway, offtopic | 15:39:57 |
| Atemu | Anyhow, Monado with STEAMVR_LH is quite good as the weak part has always been libsurvive | 15:40:45 |
| Atemu | I have migrated to it (+xrizer/opencomposite) for VR purposes and will be retiring the kernel patch from my config soon because I no longer have a need for SteamVR as a compositor | 15:42:10 |
 K900 | You need to have CAP_SYS_NICE | 15:44:28 |
| Atemu | static int amdgpu_ctx_priority_permit(struct drm_file *filp,
enum drm_sched_priority priority)
{
/* NORMAL and below are accessible by everyone */
if (priority <= DRM_SCHED_PRIORITY_NORMAL)
return 0;
if (capable(CAP_SYS_NICE))
return 0;
if (drm_is_current_master(filp))
return 0;
return -EACCES;
}
| 15:45:33 |
| K900 | OK so there's no check for root namespace | 15:46:32 |
| K900 | So technically inherited CAP_SYS_NICE would work probably | 15:46:40 |
| Atemu | bool capable(int cap)
{
return ns_capable(&init_user_ns, cap);
}
| 15:47:55 |
| Atemu | (It's really quite remarkable how readable kernel sources are despite being c) | 15:48:46 |
| K900 | Of course | 15:52:23 |
| K900 | Ugh | 15:52:31 |
| K900 | Stupid | 15:52:36 |
| K900 | Then I guess we have to wait for real DRM scheduling to become an actual thing | 15:53:01 |
| K900 | Or I guess for GPUs to grow enough hardware scheduling to make this not a problem | 15:53:42 |
| magic_rb | What, the way im reading the code, which is probably wrong, would mean that were good. It says that the init_user_ns must be capable of sys_nice, which it would be no? | 16:07:02 |
| K900 | No, it says the current thread needs to have that capability in the init ns | 16:07:38 |
| magic_rb | Oooh | 16:07:51 |
| magic_rb | Oh so its checking the capability of the thread in the root ns, not the threads ns | 16:08:15 |
| magic_rb | (I mean thats an easy patch :P) | 16:08:21 |
| magic_rb | (Would also solve the gamescope RT issue) | 16:08:31 |
| K900 | It won't no | 16:08:40 |
| K900 | That's a different issue | 16:08:44 |
| magic_rb | Is it? If gamescope had valid cap_sys_nice when launched from steam as custom command, it could renice itself | 16:09:18 |
