#asahi-dev · 2022-08-05 11:07
mazbut also, why mix this with sleep control?#asahi-dev · 2022-08-05 11:02
mazthe flip is really odd. it must be some sort of state change acknowledgement, but there are way simpler ways to do that...#asahi-dev · 2022-08-05 11:00
mazidle speculation...#asahi-dev · 2022-08-05 10:59
mazmaybe that's to prevent opportunistic sleep on the BT end. flip it to 1 to make sure it is alive, and 2 forces a sleep?#asahi-dev · 2022-08-05 10:57
mazcould that be some sort of mutual exclusion signalling between the two ends?#asahi-dev · 2022-08-02 12:42
mazthe best kind!#asahi-dev · 2022-08-02 12:42
mazI think j`ey had a vague prototype.#asahi-dev · 2022-08-02 12:23
mazmarcan: most likely direct. at this stage, Rust is its own subsystem...#asahi-dev · 2022-08-02 11:27
mazLinus was keen a few weeks ago, and I guess that from a marketing PoV, saying "Linux 6.0 supports Rust" is tempting.#asahi-dev · 2022-07-30 12:49
mazah, bonus points for the internal power supply...#asahi-dev · 2022-07-30 11:07
mazmarcan: thanks for that. I'll have a play at some point, though I fear I will need to give the mini a macos upgrade (still on the 11.something that came with it 18 months ago), and I'm willing to bet that crsutil won't work the same way there, if at all.#asahi-dev · 2022-07-29 09:32
mazsven: I'm sure it has mellowed down over time... ;-)#asahi-dev · 2022-07-29 08:22
mazthey probably saved a handful of gates by doing so.#asahi-dev · 2022-07-29 08:21
mazthanks!#asahi-dev · 2022-07-29 08:19
mazOK. let me know if I can help, I have a vested interest in having it working! ;-)#asahi-dev · 2022-07-29 08:18
mazmarcan: btw, do you have any write-up of what is needed to write the USB FW uploader on the Studio?#asahi-dev · 2022-07-22 10:42
mazhypervisors are crap. let's go shopping! :D#asahi-dev · 2022-07-22 10:42
mazyeah, I'm not actively looking, too many things going on ATM.#asahi-dev · 2022-07-22 10:41
mazbut it is really hard to trigger (8 parallel VMs on a mini, and only one failure per 6 or 7 runs)#asahi-dev · 2022-07-22 10:38
mazand only M1 triggers it.#asahi-dev · 2022-07-22 10:37
mazyup, doesn't seem unlikely. I've been chasing a similar issue with QEMU's implementation of virtio that goes in the weeds when the machine gets properly hammered.#asahi-dev · 2022-07-22 10:34
maz(though it took me some time to convince myself)#asahi-dev · 2022-07-22 10:34
maznote that this doesn't by itself order with the sysreg access that generates the IPI, but the control dependency should be enough here.#asahi-dev · 2022-07-22 10:33
mazmarcan: ISTR that we had that discussion at the time, and came to the conclusion that the atomics were enough to create the ordering.#asahi-dev · 2022-07-15 10:18
mazchadmed: try and fit AIC in ACPI first...#asahi-dev · 2022-07-13 08:34
mazj`ey: the notion of workstation is also a bit odd. I use the work laptop as a glorified X server, and use Xpra to handle sessions on both the Studio and the x86 box. not sure which of the 3 qualifies as a workstation... ;-)#asahi-dev · 2022-07-13 08:30
mazj`ey: not quite. I still have an x86 box to run the FVP.#asahi-dev · 2022-07-13 08:26
mazmoving from the D05 to a Studio has made a visible dip in my power consumption. I fear the provider is about to go bust any minute! ;-)#asahi-dev · 2022-07-12 10:56
mazBTW, if anyone has recommendations for a good USB-SATA enclosure that would take 4 3.5 HDs, I'm all ears.#asahi-dev · 2022-07-12 10:54
mazmarcan: I'm halfway there too. need to move the backups over to the Studio.#asahi-dev · 2022-07-01 11:41
mazanyway, -SIGLUNCH.#asahi-dev · 2022-07-01 11:40
mazpeople do (we certainly do), but only QEMU has this behaviour, and Linux in general isn't too affected by the CPU migration under its feet... but it's mostly luck ATM.#asahi-dev · 2022-07-01 11:38
mazdon't under-estimate the capacity of small core designers to fuck things up...#asahi-dev · 2022-07-01 11:38
mazyup. we'd tell the guest: "this is a fake CPU, but here's the list of MIDRs you may see, go and enable all the errata that match this@.#asahi-dev · 2022-07-01 11:37
mazit would also require some pretty heavy guest buy-in.#asahi-dev · 2022-07-01 11:36
mazbut that's pretty low on my laudry list.#asahi-dev · 2022-07-01 11:36
mazI have a vague plan to allow the exposure of a virtual MIDR with a list of actual cores you can run on, and let the guest deal with that.#asahi-dev · 2022-07-01 11:36
mazno. we explicitely expose the underlying HW because we have so many bugs to mitigate in the guest...#asahi-dev · 2022-07-01 11:35
maz(effectively, a VM startup is implemented as a VM migration)#asahi-dev · 2022-07-01 11:35
mazif you have two types, it goes tits up.#asahi-dev · 2022-07-01 11:35
mazno, the issue is that QEMU starts a guest by sampling the id registers on one CPU, and tries to restore them on all of them.#asahi-dev · 2022-07-01 11:34
mazkvmtool doesn't care, for example.#asahi-dev · 2022-07-01 11:34
mazmarcan: this is a QEMU issue.\#asahi-dev · 2022-07-01 11:31
mazXpra has the option to route audio. never tried though...#asahi-dev · 2022-06-27 08:33
mazsven: do we have a way to reconstruct this 'unlock key' in m1n1? it seems to be built from the device name, but *which* name?#asahi-dev · 2022-06-27 08:08
mazotherwise I'd see some drops, even with an arduino-based hack.#asahi-dev · 2022-06-27 08:07
mazsurely, on something like a mini, this should be a pretty long time.#asahi-dev · 2022-06-27 08:05
mazsven: really? damn. then pcc definitely has a point, and we should really try and do that.#asahi-dev · 2022-06-27 07:31
mazlack of coffee, let me rephrase that: I don't know of any control on the AP side to control the UART mux.