#asahi-dev · 2023-08-20 09:28
mazbrain. weird machine.#asahi-dev · 2023-08-20 09:26
mazwow. steady hand!#asahi-dev · 2023-08-20 09:25
mazthe shell isn't grounded though.#asahi-dev · 2023-08-20 09:25
mazit should be connected. it is the ID and VBUS lines that are not.#asahi-dev · 2023-08-20 09:22
mazsven: the D+/D- lines are also routed to test points right next to the uUSB connector.#asahi-dev · 2023-08-19 17:56
mazyup. the level of QC is... terrible. on that batch, I have components that are literally off by a mm... they just give you a coupon to respin it.#asahi-dev · 2023-08-19 17:51
mazI've had a number of quality issues with JLCPCB -- one batch of CS (v3) was a total dud...#asahi-dev · 2023-08-19 17:50
mazno hub involved. but a sample of one doesn't mean much.#asahi-dev · 2023-08-19 17:50
maz(i.e. it booted fine).#asahi-dev · 2023-08-19 17:49
mazs/full/high/#asahi-dev · 2023-08-19 17:49
mazFTR, I just booted a machine with the USB pass-through directly attached to the server, and it went into full speed without any issue.#asahi-dev · 2023-08-19 17:48
mazoh, interesting. thanks for the pointer. I always wondered how much I could get away with not caring too much about that, but I guess I now know... ;-)#asahi-dev · 2023-08-19 17:41
mazexpert*#asahi-dev · 2023-08-19 17:41
mazhey, you raised the issue. you're the expecrt for now! :D#asahi-dev · 2023-08-19 17:39
mazOK, thanks. I'll see if I can learn how to route these signals in a better way.#asahi-dev · 2023-08-19 17:38
maz2 layers only.#asahi-dev · 2023-08-19 17:38
mazcould you inspect the solder pads on both ends? I can also post you a more recent version, to work out whether it is something that magically improved.#asahi-dev · 2023-08-19 17:35
mazcomponents*#asahi-dev · 2023-08-19 17:35
mazI'm definitely going through a hub too (too many machines connected to the console server). given that the version you have has no active compotents between the two ends of the D+/D- lines, it must be related to the dodgy impedance of the tracks... :-/#asahi-dev · 2023-08-19 16:44
mazhas it worked before? or is it the first time you're trying to make use of it?#asahi-dev · 2023-08-19 16:43
mazsven: odd. I'm pretty sure the impedance matching of the USB lines is... iffy, but it always worked on the 3 machines I use that directly boot over the USB pass-through.#asahi-dev · 2023-08-07 08:07
mazmarcan: as it turns out, Akihiko fixed QEMU already (20230727073134.134102-1-akihiko.odaki@daynix.com) and Peter merged it for 8.2. Should be easy enough to backport to something older as required.#asahi-dev · 2023-08-07 07:59
mazmarcan: it would imply fixing libvirt, and I'm not touching that with a barge pole. but I'll have a stab at fixing QEMU if nobody has posted a patch just yet.#asahi-dev · 2023-08-06 11:15
mazfrankly, the '-M none' thing should be killed altogether. it's a lot of nonsense.#asahi-dev · 2023-07-27 09:23
mazI hope firecracker has improved its virtio support on arm64. 3 years ago, they had no idea what a memory barrier was...#asahi-dev · 2023-07-24 11:45
maz(or rather cannot support without SW involvement).#asahi-dev · 2023-07-24 11:44
mazyou seem to be missing that PSTATE is saved in SPSR during an exception entry, making (b) irrelevant. (a) cannot happen, as you'd otherwise could get nested exceptions, which the architecture can't support.#asahi-dev · 2023-07-24 11:41
maz(R_WTXBY being the rule relevant to your question)#asahi-dev · 2023-07-24 11:40
mazsid-linux: see D1.3.2 in the ARM ARM for all the gory details.#asahi-dev · 2023-07-24 11:39
mazsid-linux: by definition, DAIF are all set to 1 on exception entry, no matter where you enter from.#asahi-dev · 2023-07-24 09:12
mazCNTVCT is private as it is influenced by CNTVOFF, which is itself per-CPU.#asahi-dev · 2023-07-24 09:11
mazthings like the system PMU horror *can* be shared.#asahi-dev · 2023-07-24 09:10
maz99% of the architected register are per-CPU.#asahi-dev · 2023-07-21 10:56
mazmarcan: send me an email in private with all the details, and I'll try to do that before I have to disappear again.#asahi-dev · 2023-07-21 10:54
mazmarcan: happy to send one your way. then you can shout at me for all the bugs I've introduced in your code! ;-)#asahi-dev · 2023-07-21 10:24
mazI have too many dev versions now. someone takes them from my desk! :D#asahi-dev · 2023-07-21 10:23
mazI'm also working on being able to directly plug it to a console server (real RS232 levels)... might be useful for server-use types.#asahi-dev · 2023-07-21 10:22
mazkettenis: I'll tag the stuff later today so that people can build it.#asahi-dev · 2023-07-21 10:14
mazCentral Scrutinizer v3.1 is alive! Just got the boards back, and a sample of 1 went through the initial smoke test without issue (including being able to use dumb cables for serial...)#asahi-dev · 2023-07-11 18:31
mazsven: we should definitely add that functionality! :D#asahi-dev · 2023-07-10 18:05
mazurgh, back to a UK keyboard...#asahi-dev · 2023-07-10 18:05
mazscratch the @never saw that@, I obviously saw it.#asahi-dev · 2023-07-10 18:05
mazsven: (just got back home after a trip to Rotterdam) never saw that -- I've had the counter thingy killing my boxes for a while, but that stopped once I upgraded my host m1n1 setup, which has a counter reset hook.#asahi-dev · 2023-07-07 08:11
mazsven: congrats on making it to kernel recipe!#asahi-dev · 2023-07-03 07:23
mazyeah, I really like the automatic logging, so that I don't have to worry about capturing stack traces. and being able to add basic scripting to it helps with some limited automation.#asahi-dev · 2023-07-03 07:20
mazthat's what I've done. except that there is a 32-port USB2 hub with all the zoo connected to it, and conserver to interact with it.#asahi-dev · 2023-07-03 07:18
maz(I toyed with the idea of implementing some basic IPMI support, but like is too short fr that)#asahi-dev · 2023-07-03 07:17
maznext thing you know, you'll setup a console server with integrated reset control.#asahi-dev · 2023-07-03 07:15
mazbox has crashed? no problem!#asahi-dev · 2023-07-03 07:15
mazI specially enjoy doing that while on the train...