#asahi · 2024-02-26 14:45
eldondevhttps://github.com/AsahiLinux/m1n1/blob/main/src/wdt.c#asahi · 2024-02-26 14:45
j`ey(not sure if it is used on kernel panic)#asahi · 2024-02-26 14:45
j`eybut yes, we do have a watchdog#asahi · 2024-02-26 14:44
j`eythe kernel doesnt support kexec when using the spin-tables way of onlining CPUs#asahi · 2024-02-26 14:44
j`eydue to CPUs staying online#asahi · 2024-02-26 14:44
j`eycant kexec at all#asahi · 2024-02-26 14:43
PaulFertserI remember that it can't fully kexec because so many aux hardware can't be properly fully re-inited. But is it true even for the PCIe controller to access NVMe? It's not like anything else is needed to save the crash info is it?#asahi · 2024-02-26 14:42
leioasahi can't kexec afaik though#asahi · 2024-02-26 14:42
PaulFertserFor this argument to be less meaningful and at least somewhat educational, can someone please tell if Asahi has a hardware watchdog that reboots on kernel panic and whether it's activated by default in Fedora? Another question is whether Fedora automatically kexecs a kernel to collect crash dump and save it for later analysis prior to automatic reboot like RHEL does?#asahi · 2024-02-26 14:42
leioto me it sounds like a 5+ minute freeze that isn't waited on#asahi · 2024-02-26 14:41
leiobut yeah, we can't really tell without further info; might be that it was meant that a specific program only "crashes"#asahi · 2024-02-26 14:41
leioif a watchdog reboots it relatively quickly, it'd be a sudden reboot instead#asahi · 2024-02-26 14:41
leioI forgot the end of OP by now already; if for the user the computer doesn't do anything, they say it's a crash#asahi · 2024-02-26 14:40
eldondevFor example, a hardware fault which actually kills the processor without linux being able to panic?#asahi · 2024-02-26 14:40
eldondevThe user reported a crash. Is it possible to have a crash without a panic?#asahi · 2024-02-26 14:39
eldondevWell, that's fair, but we don't actually have information about the technical correctness of the terms AFAICT.#asahi · 2024-02-26 14:38
leiothe whole argument was due to semantics and somewhat arguing that it isn't as good as a completely freeze for users, while one side argued that it's bad while the other just wanted technical correctness of the terms :)#asahi · 2024-02-26 14:37
leioon that one, I would say 60% might easily be normal if whatever is in those falkon tabs slurps memory; It doesn't sound like cache is counted or it'd have been a 95% RAM report, not 60% (except when it was a fresh boot still)#asahi · 2024-02-26 14:37
eldondevTrue#asahi · 2024-02-26 14:36
PaulFertserEither way sospi isn't here to clarify and guessing what was meant is moot.#asahi · 2024-02-26 14:34
PaulFertsereldondev: we were discussing sospi's report where it's clear the RAM usage was monitored by operator.#asahi · 2024-02-26 14:34
leioI call it a freeze#asahi · 2024-02-26 14:33
PaulFertserMost PMU ICs will not reboot when someone hold power button no matter for how long. Holding power button usually turns the host fully off eventually but doesn't power it back unless you release and press again.#asahi · 2024-02-26 14:33
eldondevBut nontechnical people, who use the term crash, often lack the tools to observe swap thrashing.#asahi · 2024-02-26 14:32
PaulFertserIf you observe "swap thrashing" scenario and you want to give it a chance to recover yes, you'd wait.#asahi · 2024-02-26 14:31
eldondevRight! Even my nontechnical family members will hold down the power button until they get a reboot in that scenario. They definitely call it a crash.#asahi · 2024-02-26 14:31
PaulFertserI think it does a lot because if you wait you'll probably have much of your essential runtime data intact.#asahi · 2024-02-26 14:31
j`eyare you expecting people to just leave it for hours in case it starts working again?#asahi · 2024-02-26 14:31
j`eyif a users computer freezes for several minutes, doesnt really matter what it's called!#asahi · 2024-02-26 14:30
PaulFertserCrash is called "panic" in Linux terminology and in that case a distro can often use some machinery to save a crash dump and then reboot (enabled by default in RHEL). Another kind of crash I can imagine is when hardware watchdog forces reboot. Freezing for minutes (and sometimes hours) due to low RAM and then coninuing to run normally is by no sane definition a crash, I'm pretty sure.#asahi · 2024-02-26 14:27
j`eythat's basically a crash for users#asahi · 2024-02-26 14:26
PaulFertserSure, freezing for minutes (and hours) can happen especially when there's swap space enabled. But it's not a crash.#asahi · 2024-02-26 14:25
PaulFertserkidplayer666: issues meaning the kernel paniced or watchdog triggered or what?#asahi · 2024-02-26 14:25
leioI believe systemd-oomd or something hopefully made it better in newer versions when enabled#asahi · 2024-02-26 14:24
leio(existing SSH sessions don't move either)#asahi · 2024-02-26 14:24
leiosome fedora amazon images I deal with also routinely freeze for 10 minutes and are unresponsive when OOM occurs before it recovers from some kills; can't SSH, can't log in from console, nothing, until eventually they wake up again after killing something; admittedly that's older than fedora 39 though.#asahi · 2024-02-26 14:23
kidplayer666cause the default one seems quite a bit oversaturated#asahi · 2024-02-26 14:23
kidplayer666also, anyone done an icc profile for the mac screens?#asahi · 2024-02-26 14:22
kidplayer666PaulFertser: I've had several issues with memory associated crashes, so i wouldn't be so sure about that#asahi · 2024-02-26 14:11
eldondevJust trying not to damage my hardware as it's the only resource I have to work with Asahi on.#asahi · 2024-02-26 14:08
eldondevSorry if I'm talking too much about this stuff or in the wrong place :(#asahi · 2024-02-26 14:07
eldondevLooks like macos supports some sort of "thunderbolt bridge," which also sounds intrigueing, but thunderbolt isn't on the support matrix yet.#asahi · 2024-02-26 14:07
eldondevAlso, the article I linked does some funky thunderbolt linking thing to link the two machines.#asahi · 2024-02-26 14:06
eldondevnicolas17: I was just wondering if I needed to do anything special regarding DTB's etc to use the dual role controller in device mode on my Mac Mini(2020), that feels like an asahi thing.#asahi · 2024-02-26 14:00
PaulFertserI would say discussing memory pressure issues makes little sense without seeing /proc/meminfo .#asahi · 2024-02-26 13:58
PaulFertserj`ey: yes, having oom killer do its job is no joy but no system crash either.#asahi · 2024-02-26 13:54
j`eysospi: can you run 'swapon --summary' do you have a swapfile, or just the zram?#asahi · 2024-02-26 13:54
j`eyif the oom killer works that is..#asahi · 2024-02-26 13:53
PaulFertsersospi: with default settings Linux always uses free mem for caching access to storage. The cache is dropped when more memory is needed. You should provide more info about your situation. Also Linux doesn't crash when it's low on RAM, it runs an oom-killer to terminate the most suitable process to free up memory.#asahi · 2024-02-26 13:29
sospiI, I'm an asahi user (M1 macbook air) and I want to know if it is normal to have high memory usage on asahi. For exemple, I'm running sway + 4 falkon tab + Hex chat and i'm at 60% ram usage. It makes quite impossible to use firefox normally (always >80% with a few tab open) and my pc regularry crash bc of this... I already had 8gb pc and never ran into this.. Thx :)