#asahi-dev · 2022-02-27 19:16
kettenisand verified that adding the nodes for disp_dart and dcp_dart has the expected effect#asahi-dev · 2022-02-27 18:59
kettenissven: that seems to work#asahi-dev · 2022-02-27 18:12
kettenisso adding a simple TCR check is not going to fly unless we have m1n1 sanitize the DARTs#asahi-dev · 2022-02-27 18:11
kettenishmm, on the j274, 4 out of the 8 DARTs that are currently in the device tree have the translate enable bit set (for all of the streams)#asahi-dev · 2022-02-27 16:01
kettenisGuess the TCR check will buy us some time for that#asahi-dev · 2022-02-27 16:00
kettenisBut I suspect this will still take a bit of time to be worked into a formal binding...#asahi-dev · 2022-02-27 16:00
kettenisI guess that would work; the presence of the mapped-devices property would then indicate that we don't want to do a full reset of the DART I presume?#asahi-dev · 2022-02-27 15:30
kettenisbut I'll see if I can do the TCR check on OpenBSD as well#asahi-dev · 2022-02-27 15:29
kettenisFWIW, I'd like to see a property that indicates that mappings need to be preserved#asahi-dev · 2022-02-27 15:27
kettenismaybe not in that case#asahi-dev · 2022-02-27 15:25
kettenis(not that I necessarily agree with that)#asahi-dev · 2022-02-27 15:24
kettenisnote that robh has indicated a few times that using different compatibles is generally preferred over boolean properties like that#asahi-dev · 2022-02-27 15:22
kettenisand have the driver key off that#asahi-dev · 2022-02-27 15:21
kettenisThe alternative solution is to change the compatible for these display DARTs to something like apple,t8103-disp-dart#asahi-dev · 2022-02-27 15:18
kettenis(and 5.18 if the change doesn't make it into that)#asahi-dev · 2022-02-27 15:17
kettenisnote that this means that 5.17 will still break#asahi-dev · 2022-02-27 15:16
kettenisthanks ;)#asahi-dev · 2022-02-27 15:12
kettenisIt probably should undo that, but it currently doesn't.#asahi-dev · 2022-02-27 15:12
kettenisThe only thing u-boot does is place some of the DARTs in bypass mode.#asahi-dev · 2022-02-26 12:52
kettenisfurther exploration of those bits will have to wait, since I definitely have a problem with non-statically assigned interrupts#asahi-dev · 2022-02-26 12:47
kettenisand only setting one of the bits doesn't#asahi-dev · 2022-02-26 12:29
kettenissetting both bits definitely has the desired effect#asahi-dev · 2022-02-26 11:43
kettenisI'll poke at it for a bit#asahi-dev · 2022-02-26 11:43
kettenisbecause I'm not entirely confident OpenBSD can handle non-statically assigned interrupts#asahi-dev · 2022-02-26 11:42
ketteniswell, I'm trying to get all interrupts delivered on the primary CPU#asahi-dev · 2022-02-26 11:40
kettenisit only turns of IRQs, not FIQs?#asahi-dev · 2022-02-26 11:39
kettenisin particular, can I use this to disable IRQs on all but the primary CPU?#asahi-dev · 2022-02-26 11:38
kettenismarcan: did you figure out a bit more about the s3_4_c15_c5_0 register that m1n1 writes to enable IRQs?#asahi-dev · 2022-02-25 21:47
kettenisbut definitely time for a beer#asahi-dev · 2022-02-25 21:44
kettenisstill need to make pcie work, and the new wireless#asahi-dev · 2022-02-25 21:33
kettenis(but I guess someone else did all the hard work)#asahi-dev · 2022-02-25 21:33
kettenisthat was too easy#asahi-dev · 2022-02-25 21:33
kettenismainbus0 at root: Apple MacBook Pro (14-inch, M1 Pro, 2021)#asahi-dev · 2022-02-25 20:38
kettenisah, good point#asahi-dev · 2022-02-25 20:26
kettenis(I never do it)#asahi-dev · 2022-02-25 20:26
kettenisjannau: seems the savedefconfig step can be ommitted when sendig diffs#asahi-dev · 2022-02-25 20:06
ketteniswe'll have to see; hopefully the asahi branch ends up as master with a few light customizations#asahi-dev · 2022-02-25 15:17
kettenisdo you need to max out the max to benefit from the higher memory bandwidth?#asahi-dev · 2022-02-25 15:03
kettenisjust in time for the weekend...#asahi-dev · 2022-02-25 15:02
kettenisjay, my m1 pro machine arrived#asahi-dev · 2022-02-25 11:56
kettenisuses special cpu registers instead of having to go through the AIC, which may be why it is so much faster#asahi-dev · 2022-02-25 11:55
kettenisdelivering inter-processor interrupts as fiq#asahi-dev · 2022-02-25 11:52
kettenis(my hypothesis is that faster IPI delivery results in less "merging" of IPIs sent by different CPUs#asahi-dev · 2022-02-25 11:51
kettenisI swithed OpenBSD to fast-ipis recently and noticed that the total number of IPIs that get delivered as a fraction of IPIs that get sent increased significantly#asahi-dev · 2022-02-23 12:06
kettenischadmed: :D#asahi-dev · 2022-02-23 12:04
kettenissven: not sure where that originates, but it is what we tend to call it in openbsd ;)#asahi-dev · 2022-02-23 12:03
kettenisif I add this code to u-boot there will be a way to override this#asahi-dev · 2022-02-23 12:02
kettenisheh#asahi-dev · 2022-02-23 12:01
kettenisso maybe I can implement that in u-boot#asahi-dev · 2022-02-23 12:01
kettenisI always liked how Sun machines used to switch from serial console to glass console when you plugged in a keyboard