#
✕ nick:sven
#asahi · 2022-09-27 12:29
svenso if you disable those effects (either because you don't have a 3d driver and they are automatically disabled I guess or just doing it manually) makes everything more ugly but also snappier
#asahi · 2022-09-27 12:28
svenadditional graphical effects sometimes introduce a delay, e.g. the slowing shrinking the window when you minimize it that macos does vs. just hiding it immediately
#asahi · 2022-09-21 15:11
svenyou can read me whine and complain about dcpext and DP altmode which is required for that hdmi out in #asahi-dev :D
#asahi · 2022-09-21 14:44
sven+see
#asahi · 2022-09-21 14:44
svenI think you can the entire history if you join it
#asahi · 2022-09-21 14:42
sven#asahi:fedora.im
#asahi · 2022-09-21 14:10
sventhere are some fedora people working on official support afaik, but I don’t think they have anything ready for public testing yet
#asahi · 2022-09-21 11:19
svendifferent level of quality assurance and you don’t get to complain here if anything breaks ;)
#asahi · 2022-09-19 19:04
svenSo I guess we need to call that parameter please-destroy-my-speakers-this-is-not-a-joke or something
#asahi · 2022-09-19 19:03
svenProgram aborted. I thought was a joke of hdparm or Linux.” ugh
#asahi · 2022-09-19 19:03
sven“Please also supply the --please-destroy-my-drive flag if you really want this.
#asahi · 2022-09-18 16:54
svenlooks like the log is incomplete for some reason
#asahi · 2022-09-17 06:20
svenI just don’t know how to ask DCP yet
#asahi · 2022-09-17 06:20
svenphire: i think I have the usb c stuff down and now just the “ask DCP nicely” is missing
#asahi · 2022-09-08 21:28
svenI still don’t understand the entire mess well enough to actually start thinking how it fits into Linux
#asahi · 2022-09-08 21:28
svenmaybe, but maybe parts of it will be separate and become a mux or something (the crossbar part)
#asahi · 2022-09-08 21:26
svenand technically the DP alt mode driver is separate from the display driver as well :D
#asahi · 2022-09-03 11:21
svenit might need thunderbolt support I guess
#asahi · 2022-09-01 14:47
svenprobably it just feeds the raw output images into a complex ML model to improve it. i'd be surprised if we can't grab that raw image more easily
#asahi · 2022-08-30 18:53
svenwhich is already complicated enough :D
#asahi · 2022-08-30 18:52
svenit's very far from done
#asahi · 2022-08-30 07:06
svenso that means thunderbolt will be delayed by three more months :>
#asahi · 2022-08-29 14:55
svenit’ll probably make stuff go slower tbh
#asahi · 2022-08-26 16:19
svenand then add the usual Apple shenanigans on top where the NHI (~ the host router) is not a pcie device for which linux has a driver but a platform device again. and ofc it's all somehow also connected to the type-c phy which has to expose the correct electrical signals
#asahi · 2022-08-26 16:18
svenyeah, pretty much
#asahi · 2022-08-26 16:18
svenand ofc that can also all be chained because it's not complicated enough yet
#asahi · 2022-08-26 16:17
sven*very thin
#asahi · 2022-08-26 16:17
svenso you can just mux all those at the host, transport them over a very thing cable and demux them on the other end and expose them on ports
#asahi · 2022-08-26 16:16
svenessentially the usb4 stream can transport pcie, usb3 and DP packets
#asahi · 2022-08-26 16:16
svenhttps://tweakers.net/ext/i/2003046178.png I think
#asahi · 2022-08-26 16:15
sventhere's a "lovely" picture in the usb4 spec, let me find that again
#asahi · 2022-08-26 16:14
svenfrom what I understand DP tunneling is essentially a DP stream encapsulated and muxed in USB4 packets. a downstream usb4 router can then just decode that stream and expose it
#asahi · 2022-08-26 16:11
svenstill need to understand dp and dp tunneling if I want to upstream this entire mess :(
#asahi · 2022-08-26 16:11
svenfair enough :D
#asahi · 2022-08-26 16:08
svenactually working on TB right now. I'll have to understand the atcphy <-> display port interaction to eventually come up with sane DT bindings :-)
#asahi · 2022-08-26 16:06
svenyeah, and I really like to use that as an excuse to slack off and ignore that 4k patch :D
#asahi · 2022-08-26 16:05
svenrobin has a WIP series that improves the API and makes that 4k patch much less awkward
#asahi · 2022-08-26 16:04
sveni think you just complained about the iommu api design as well :D
#asahi · 2022-08-26 16:02
sventhe iommu currently assumes that there's a single type of IOMMU for each bus and you don't really know much about an iommu domain before the first device is attached
#asahi · 2022-08-26 16:01
svenand robin's patch series fixes that. so once that lands i'll re-spin that 4k one
#asahi · 2022-08-26 16:00
sveniirc it was one of the iommu guys. the problem really is that the IOMMU api is a bit awkward for this case
#asahi · 2022-08-26 15:59
svenhttps://gitlab.arm.com/linux-arm/linux-rm/-/commits/iommu/bus that thing makes the 4k patchset less awkward
#asahi · 2022-08-26 15:59
svenlet me find robin's cleanup again
#asahi · 2022-08-26 15:58
svenhttps://lore.kernel.org/linux-iommu/20211019163737.46269-1-sven@svenpeter.dev/ is that patch series (which probably doesn't apply anymore)
#asahi · 2022-08-26 15:57
sveni have a WIP patchset somewhere that teaches the iommu subsystem to handle that case but i'm waiting for a cleanup series of robin to land before submitting that one again
#asahi · 2022-08-26 15:57
sventhat would also help I guess
#asahi · 2022-08-26 15:56
sventl;dr: the iommu has a hardcoded 16k pagesize and the iommu subsystem currently doesn't handle the cpu page size < iommu page size case
#asahi · 2022-08-26 15:55
sveni really need to make that error louder in the iommu driver
#asahi · 2022-08-26 15:55
sventhat explains it, it can't allocate DMA buffers with 4k kernels
#asahi · 2022-08-26 15:55
svenoh...

plus ›