#asahi · 2022-12-04 05:14
marcanyeah, I know#asahi · 2022-12-04 05:14
marcanI don't think it violates ordering guarantees, but I'm not sure anyone really stress-tested that#asahi · 2022-12-04 05:13
marcan(up to one second, but still)#asahi · 2022-12-04 05:13
marcanwithout this we still have that corner case where if your laptop runs out of battery you lose ostensibly committed data#asahi · 2022-12-04 05:13
marcanin that case it makes sense for the kernel to flush NVMe, yes, because that completes the safety scorecard for the NVMe deferred flush story#asahi · 2022-12-04 05:12
marcanthe property goes high like 10 seconds before hard shutdown or something#asahi · 2022-12-04 05:12
marcanthis is about last-gasp panic shutdown#asahi · 2022-12-04 05:12
marcanbut obviously that in and of itself wastes some power, so I'm not sure if it's a good idea (and it's not implemented yet anyway)#asahi · 2022-12-04 05:12
marcanso my idea was to just have the battery driver poll that every couple of seconds or something when the battery % is < 3% or something like that, and do a panic shutdown when it sees it#asahi · 2022-12-04 05:11
marcanwhat I did find is an SMC property that flips on#asahi · 2022-12-04 05:11
marcanI looked into the kernel side (at least flushing NVMe) but I didn't find an SMC event for it#asahi · 2022-12-04 05:10
marcanthat's up to userspace#asahi · 2022-12-02 01:27
marcanBcrumb: Because we can't ship it in the root, because it makes cleanup easier, because it makes root backups redistributable for people, and because that way it works without mounting the root read-write in the initramfs (which a lot of distros don't)#asahi · 2022-12-01 02:48
marcanlina said even with that patch vsync (in terms of framerate-limiting) is broken on xorg, and there's probably no real way of fixing it, so it's not a complete solution#asahi · 2022-12-01 02:47
marcanzx2c4: If it's going upstream, no need to push for it. People who want a tear-free experience now can try wayland ;)#asahi · 2022-11-29 15:18
marcanand that was just an unrelated issue#asahi · 2022-11-29 15:18
marcanbcrumb: so your problem probably has nothing to do with the i2c issue/probe deferral thing you saw then#asahi · 2022-11-29 07:17
marcanbcrumb: I pushed an update to asahi-scripts to asahi-dev which might help you out. switch your repo to that in /etc/pacman.conf and let me know how it goes (if it works, switch it back, you don't want to stay on asahi-dev and it won't downgrade you)#asahi · 2022-11-29 06:07
marcanah, this should be #-dev#asahi · 2022-11-29 06:07
marcan(not sure if we need an explicit ack from rob/K for the initial one? K already reviewed v3 and only had one trivial comment anyway and the binding is reviewed so possibly not)#asahi · 2022-11-29 06:07
marcanI'll put together t600x/t8112 DT patches and send them off, just need an ack from someone here and then I think we're good for merging them in principle#asahi · 2022-11-29 06:06
marcankettenis_: the cpufreq binding is reviewed, so that should be good to go as far as your side goes#asahi · 2022-11-29 06:04
marcanI can escalate#asahi · 2022-11-29 06:04
marcanif it keeps happening a day or so later though let me know#asahi · 2022-11-29 06:04
marcanthat usually means more stuff is broken and someone should be fixing it :p#asahi · 2022-11-29 06:03
marcanprobably broken intertubes between whatever POP you use and our backend server, not much we can do about that#asahi · 2022-11-29 06:02
marcansounds like a likely routing issue#asahi · 2022-11-29 02:56
marcanbcrumb: that's a build time script that adds modules, it can't "hang" on that script because it doesn't actually have anything to do with boot time#asahi · 2022-11-28 19:31
marcanclassic synopsys#asahi · 2022-11-28 19:30
marcanand suspending devices too, apparently...#asahi · 2022-11-28 19:29
marcanaxboe: yeah, try enabling all the gadget stuff and dual role#asahi · 2022-11-28 19:28
marcanand the race condition is related maybe?#asahi · 2022-11-28 19:28
marcanyeah that.#asahi · 2022-11-28 19:28
marcanalso make sure CONFIG_USB_DWC3_DUAL_ROLE=y#asahi · 2022-11-28 18:49
marcan rmdir asmedia brcm 2>/dev/null || true#asahi · 2022-11-28 18:49
marcan awk '{ print $2 }' < .vendorfw.manifest | xargs rm -f#asahi · 2022-11-28 18:49
marcan printf '==> Removing obsolete vendor firmware from /lib/firmware...\n'#asahi · 2022-11-28 18:49
marcanthe asahi-scripts postupgrade does that#asahi · 2022-11-28 18:48
marcanso if you have the old firmware mixed in you should delete that. look at the manifest.txt file in /boot/efi/vendorfw and delete any files in that list in /lib/firmware/ (directly in the root, not under vendor)#asahi · 2022-11-28 18:48
marcanhowever, the old scripts put the firmware outside of vendor and that is, if nothing else, ugly, though it could actually break your wifi in the future under rare but known to happen circumstances#asahi · 2022-11-28 18:47
marcanthere is no real side effect to leaving stuff in vendor since the initramfs will mount a tmpfs on top anyway#asahi · 2022-11-28 18:22
marcanif your initramfs stuff is not working then just untar the firmware.tar into /lib/firmware/vendor (and then fix it so it works and clean up that directory to undo it, because that solution is partial and runs into all the race condition problems we've been having, e.g. probably won't work for the xhci controller firmware)#asahi · 2022-11-28 18:21
marcananyway please don't ever downgrade asahi-scripts/asahi-fwextract/etc, we don't test that path and it's just going to make everything more confusing#asahi · 2022-11-28 18:19
marcangot two users saying their wifi broke after the upgrade already, 99% sure they uninstalled asahi-meta/asahi-fwextract at some point...#asahi · 2022-11-28 18:18
marcansighs, why do people try so hard to break their systems in weird ways :p#asahi · 2022-11-28 18:18
marcanand that the correct thing to do would've been to just untar /boot/efi/vendorfw/firmware.tar into /lib/firmware/vendor as a stopgap until the initramfs stuff is working and mounting a tmpfs there instead...#asahi · 2022-11-28 18:17
marcansomeone tell bcrumb when they're back that the old firmware put stuff in /lib/firmware directly and the upgrade process is designed to clean it up because it now goes in /lib/firmware/vendor but they probably ended up in a weird state with the downgrades/upgrades and now the old firmware is still there so it "works" (but won't work on upgrades)#asahi · 2022-11-24 17:14
marcan(same thing happens when Apple's update stuff goes wrong or you don't do step2 to begin with)#asahi · 2022-11-24 17:14
marcanwhich will detect what you just did as broken#asahi · 2022-11-24 17:13
marcanrather you can use the installer's "fix my install" mode