#
✕ nick:sven
#asahi-dev · 2025-08-24 15:38
svenyeah, i just forgot to run it before i submitted that rfc
#asahi-dev · 2025-08-24 15:34
sveni've used make dt_binding_check DT_SCHEMA_FILES=Documentation/devicetree/bindings/phy/apple,atcphy.yaml and fixed the warnings from there
#asahi-dev · 2025-08-24 15:16
svensame branch as above now has (probably) fixed dt-bindings
#asahi-dev · 2025-08-24 14:37
sven*there‘s
#asahi-dev · 2025-08-24 14:37
sventhere nothing in the Linux driver that depends on the compatible now that DP-only is gone
#asahi-dev · 2025-08-24 14:36
svenfrom what I can tell the compatible is different in ADT because the driver hardcoded the fuses
#asahi-dev · 2025-08-24 14:35
svenand we have a bit better control over how the shutdown/reinit actually happens during role switches instead of hacking dwc3_set_mode
#asahi-dev · 2025-08-24 14:25
svenright now we have no fallback there otherwise. should we also do "apple,t6000-atcphy", "apple,t8103-atcphy" there? i think apple hides incompatibilities between these in the tunables
#asahi-dev · 2025-08-24 14:24
sveni'm gonna drop the dp-only compatible for the atcphy
#asahi-dev · 2025-08-24 14:02
svenhttps://git.kernel.org/pub/scm/linux/kernel/git/sven/linux.git/log/?h=wip/atcphy it's starting to look reasonable again
#asahi-dev · 2025-08-24 13:46
svenso that's another core commit we don't need anymore
#asahi-dev · 2025-08-24 13:46
svenjust played around with the susphy stuff, looks like it's only required just before calling dwc3_core_exit
#asahi-dev · 2025-08-24 13:46
svennice
#asahi-dev · 2025-08-23 19:15
sveni'm really not a big fan of dropping snps,dwc3 but i haven't been able to make it work otherwise fwiw
#asahi-dev · 2025-08-23 16:08
svenhaven't updated the dt bindings yet
#asahi-dev · 2025-08-23 16:08
svenhttps://git.kernel.org/pub/scm/linux/kernel/git/sven/linux.git/log/?h=wip/atcphy& current status
#asahi-dev · 2025-08-23 15:47
svendunno why that makes the pipehandler lock fail though
#asahi-dev · 2025-08-23 15:47
svenhrm, i still need that susphy thing
#asahi-dev · 2025-08-23 13:14
svenyeah, having snps,dwc3 together with the glue.h thing is a bad idea
#asahi-dev · 2025-08-23 13:07
sven*true
#asahi-dev · 2025-08-23 13:06
svenugh, trure
#asahi-dev · 2025-08-23 13:06
sventhat's what i'm hoping for
#asahi-dev · 2025-08-23 13:05
svenoh, yes, definitely
#asahi-dev · 2025-08-23 13:04
sven*selects
#asahi-dev · 2025-08-23 13:04
svenlet me try to find the code that actually select the driver to probe and see what would happen if leave snps,dwc3 in there
#asahi-dev · 2025-08-23 13:02
svenit’s just that losing snps,dwc3 will break uboot and openbsd
#asahi-dev · 2025-08-23 13:02
sventhe node itself looks the same for the glue driver otherwise
#asahi-dev · 2025-08-23 13:01
svenyes
#asahi-dev · 2025-08-23 12:59
svenI could make the dt-bindings work but I don’t think having two drivers compete over a single node is a good idea
#asahi-dev · 2025-08-23 12:59
svenyeah, that’s the issue
#asahi-dev · 2025-08-23 12:30
svenonly annoying part with the glue approach is that I can’t use snps,dwc3 as fallback anymore
#asahi-dev · 2025-08-23 11:56
svenhttps://f.svpe.de/9a8f20f7508956fbc06d3807a5da18014da9dd909156ce7ef96d4d86df941fff_0001-usb-dwc3-Use-ioremap_np-when-required-in-dwc3_power_.patch that fixes it
#asahi-dev · 2025-08-23 11:44
sven*fixes
#asahi-dev · 2025-08-23 11:44
svenyup, that fices it
#asahi-dev · 2025-08-23 11:35
svennope :D it's mapping the MMIO with the wrong attribute afaict
#asahi-dev · 2025-08-23 11:33
svenuh, does ioremap take nonposted into account?
#asahi-dev · 2025-08-23 11:28
sveni'm not sure why though, need to take a closer look
#asahi-dev · 2025-08-23 11:28
svenstill serrors without that host-vbus-glitch stuff
#asahi-dev · 2025-08-23 11:27
svenand the vendor register stuff
#asahi-dev · 2025-08-23 11:27
svenit'll at least drop the role switch reset quirk
#asahi-dev · 2025-08-23 11:17
svenphew, looks like it's working again
#asahi-dev · 2025-08-23 11:17
sven[ 153.090993] usb 4-1: new SuperSpeed USB device number 2 using xhci-hcd
#asahi-dev · 2025-08-23 09:01
svenyeah, the code looks much nicer now but it doesn't work anymore :D
#asahi-dev · 2025-08-23 08:56
svenhere we go again
#asahi-dev · 2025-08-23 08:56
sven[ 18.955867] phy-apple-atc 383000000.phy: pipehandler lock not acked *sigh*
#asahi-dev · 2025-08-22 14:41
svenjust started prototyping with glue.h and this already looks much nicer so far
#asahi-dev · 2025-08-22 09:26
svenand when I originally reverse engineered sprr that just wasn’t enabled in el1 vms under macOS unless you had a special apple entitlement
#asahi-dev · 2025-08-22 09:25
svenuh, gl1 require sprr in el1 to be enabled and that would really mess any Linux vm up
#asahi-dev · 2025-08-22 08:31
sventalking about el1 here
#asahi-dev · 2025-08-22 08:30
svenso maybe if can drop to el1 and disable SPRR/GXF there then and map all of physical memory

plus ›