#asahi-dev · 2023-01-30 17:00
mazhow long will it take to build defconfig on 4 M2 e-cores in a L2 guest? I'm taking bets.#asahi-dev · 2023-01-28 11:54
mazbut*#asahi-dev · 2023-01-28 11:54
mazyeah, I'd love Robin to take a more active maintainer role, be he knows better! :D#asahi-dev · 2023-01-28 11:53
mazor even ask him to drop the patch altogether.#asahi-dev · 2023-01-28 11:52
mazyou can send a revert, I'm sure he'll take it.#asahi-dev · 2023-01-28 11:06
mazyou're a wicked, wicked man!#asahi-dev · 2023-01-28 10:54
maztoo bad it is too early for a beer!#asahi-dev · 2023-01-28 10:53
mazand with that, EFI works in a L2 guest, and the Debian install ran to completion.#asahi-dev · 2023-01-28 10:48
mazfunny how things work better when you handle S2 invalidation *with the correct MMU context*...#asahi-dev · 2023-01-26 19:58
mazChampagne!#asahi-dev · 2023-01-26 14:10
mazthanks!#asahi-dev · 2023-01-26 14:09
mazmarcan: ^^#asahi-dev · 2023-01-26 14:09
mazplease report this sort of stuff on the list, cc'ing null, Sudeep Holla, and maybe Saravana Kannan...#asahi-dev · 2023-01-25 11:13
mazit did! and Cheepnis is an awesome tune!#asahi-dev · 2023-01-25 11:10
mazquote*#asahi-dev · 2023-01-25 11:10
mazdid someone just quite FZ here?#asahi-dev · 2023-01-24 14:40
mazI'm >that< close to send the design to the fab...#asahi-dev · 2023-01-24 14:39
mazit'd better work!#asahi-dev · 2023-01-24 08:56
mazthe curse is the connector, I swear.#asahi-dev · 2023-01-24 08:56
mazreset the NHI? and all that time you were claiming that it was way better than dwc3... ;-)#asahi-dev · 2023-01-24 08:53
mazoh, absolutely. There is hardly a case for HP without getting the rest of TB in shape.#asahi-dev · 2023-01-24 08:49
mazinteresting. yet it has interrupts to signal link-up/down... i guess they settled on a hack rather than supporting the full HP spec...#asahi-dev · 2023-01-24 08:36
mazI would have hoped that the same mechanism would work :-/#asahi-dev · 2023-01-24 08:31
mazhuh, how did it work the first place?#asahi-dev · 2023-01-23 17:14
mazah, so that's really the routing of the HP interrupt that fails.#asahi-dev · 2023-01-23 17:12
mazisn't this one of the complicated setups where you can't establish an end-to-end tunnel, but need to instruct the "port-in-the-middle" to do something fancy?#asahi-dev · 2023-01-23 15:44
mazmissing synchronisation between the PCIe driver and the dart on unplug?#asahi-dev · 2023-01-20 13:52
mazmakes you wonder what sort of glue they used to put this contraption together...#asahi-dev · 2023-01-20 13:50
mazsven: tg3 is a horrible driver (to be fair, most bcm drivers are like that). the -1 returns are pretty worrying though.#asahi-dev · 2023-01-20 13:48
mazI wonder if there is some aggressive PM going on on the TB side that would delay something...#asahi-dev · 2023-01-20 13:47
mazsven: do you get similarly high latency with local communication?#asahi-dev · 2023-01-20 12:36
mazright. it could be that the MSI write is buffered somewhere. can you configure the memory type for the MSI mapping?#asahi-dev · 2023-01-20 12:34
mazare you doing MSIs? or INTx?#asahi-dev · 2023-01-20 12:34
mazfeels like interrupt latency, not DMA.#asahi-dev · 2023-01-19 12:44
mazyup. and I'll probably start looking at VFIO support once TB hits upstream.#asahi-dev · 2023-01-19 12:24
mazyou wouldn't expose the IOMMU to the guest, just control it from the host.#asahi-dev · 2023-01-19 12:21
mazbypass on externally accessible ports... madness.#asahi-dev · 2023-01-18 15:37
mazbeautiful! #asahi-dev · 2023-01-17 14:46
mazthe PCB would solve that problem.#asahi-dev · 2023-01-17 14:46
mazthe expensive part here was the QFN adapter being delivered from Canada.#asahi-dev · 2023-01-17 14:45
mazwhat I like about it is that the pico is the serial adapter. and in theory, one adapter could deal with two FUSB302 (one per I2C bus), which is useful when you have multiple boxes.#asahi-dev · 2023-01-17 14:43
mazkettenis: this is what it looks like: http://www.loen.fr/tmp/PXL_20230115_121106090.MP.jpg#asahi-dev · 2023-01-17 14:38
mazI have half of a proper PCB designed, but need time to finish it...#asahi-dev · 2023-01-17 14:37
mazRPi Pico, FUSB302 on an adapter board, various level shifters for the uart and VBUS.#asahi-dev · 2023-01-17 14:36
maztime to start building a 4th serial dongle, just in case...#asahi-dev · 2023-01-17 14:33
mazwhat I'd really like is a mask on the vgic maintenance interrupt. can't imagine that happening either (and I wouldn't know where to look)#asahi-dev · 2023-01-17 14:31
mazthough TBH, it isn't too hard to work around.#asahi-dev · 2023-01-17 14:31
mazhere's to hope!#asahi-dev · 2023-01-17 14:20
mazstarts looking at ebay for a second-hand M1 mini...