#
✕ nick:marcan
#asahi-dev · 2021-11-21 22:11
marcanheh
#asahi-dev · 2021-11-21 22:11
marcanoh wait, I do lol
#asahi-dev · 2021-11-21 22:11
marcanI don't get that though :/
#asahi-dev · 2021-11-21 22:10
marcanthat's saying there's a missing parent relationship or so
#asahi-dev · 2021-11-21 22:10
marcanunrelated (probably) but that dispdfr_be thing is interesting
#asahi-dev · 2021-11-21 22:07
marcanalso did you compile in the PMGR driver? :)
#asahi-dev · 2021-11-21 22:07
marcandmesg too
#asahi-dev · 2021-11-21 22:07
marcancan you paste it somewhere?
#asahi-dev · 2021-11-21 22:07
marcanstill no SError without a display
#asahi-dev · 2021-11-21 21:59
marcanactually I'm testing in the hypervisor, haven't tested bare metal
#asahi-dev · 2021-11-21 21:59
marcanit does; that sounds like a power domain not set properly?
#asahi-dev · 2021-11-21 21:52
marcankettenis: yeah, I know, I'll rebase that anyway; I'm just using this branch as my kitchen sink to exercise the patches
#asahi-dev · 2021-11-21 16:05
marcanthe max speed thing is tangential, it defaults to Gen4 but still comes up at Gen1 without that bit
#asahi-dev · 2021-11-21 16:04
marcanbecause without PORT_LOGIC_SPEED_CHANGE preset (can be done before initial link training), it just comes up at Gen1
#asahi-dev · 2021-11-21 16:04
marcanyou get this: [ 0.266469] pci 0000:02:00.0: 2.000 Gb/s available PCIe bandwidth, limited by 2.5 GT/s PCIe x1 link at 0000:00:01.0 (capable of 4.000 Gb/s with 5.0 GT/s PCIe x1 link)
#asahi-dev · 2021-11-21 16:03
marcannot this one though
#asahi-dev · 2021-11-21 16:00
marcanalso I'm outright adding the entire t8103 pmgr tree instead of just a subset
#asahi-dev · 2021-11-21 16:00
marcanmain changes are I enable auto-PM now (unconditionally, seems to be fine) and also force on "always-on" things that aren't, but complain. There's also a bit of cleanup/fixes, and also the t6000 compatible in the bindings.
#asahi-dev · 2021-11-21 15:59
marcanAlso, the pmgr stuff in t6000/bringup-work is what I'm going to send as v3, modulo I need to do a final check to make sure I didn't do anything silly :)
#asahi-dev · 2021-11-21 15:57
marcanI kind of hate that. It's poking config space registers, but since this is before the PCI bus gets initialized, we can't use the usual accessors...
#asahi-dev · 2021-11-21 15:57
marcanhttps://github.com/AsahiLinux/linux/commit/35ee181fdfdea4c5c87b2365c6f3de707ddef745
#asahi-dev · 2021-11-21 15:06
marcan(that codebase does the speed dance properly that apple-pcie doesn't)
#asahi-dev · 2021-11-21 15:05
marcanremind me again why we aren't reusing any code from drivers/pci/controller/dwc? It's the same registers for the parts that are shared...
#asahi-dev · 2021-11-21 15:05
marcanalso there's some extra stuff that has to happen with the designware registers to make it train at the right speed on the first go (ASPM ends up making it retrain anyway, but that's after linux complains about a speed bottleneck)
#asahi-dev · 2021-11-21 15:04
marcanmaz: so in apple-pcie, we don't actually do anything with the max link speed
#asahi-dev · 2021-11-20 12:44
marcanI just figured whatever ANS firmware they used for it did things differently
#asahi-dev · 2021-11-20 12:43
marcanit doesn't really make any sense for T2 to emulate another layer on top of ANS, it's already NVMe
#asahi-dev · 2021-11-20 12:42
marcanbut linear SQ being an option makes sense in that context, yeah
#asahi-dev · 2021-11-20 12:42
marcanalways assumed it was ANS passed through
#asahi-dev · 2021-11-20 12:42
marcansven: I never thought T2 emulates NVMe
#asahi-dev · 2021-11-15 16:33
marcanokay yeah, pretty sure it was the SPMI interfaces and indeed it was crashing SMC
#asahi-dev · 2021-11-15 16:10
marcanISTR sven might've seen this before?
#asahi-dev · 2021-11-15 16:10
marcanmethinks I'm shutting down something that breaks the SMC or similar, and that causes a panic later
#asahi-dev · 2021-11-15 16:09
marcanand now it boots, except it randomly reboots
#asahi-dev · 2021-11-15 16:09
marcanso I added all the pmgr pwrstates to the DT, and set the ones that obviously break things to always on...
#asahi-dev · 2021-11-15 13:48
marcanpushed a change to delete the nodes
#asahi-dev · 2021-11-15 13:35
marcanlovely
#asahi-dev · 2021-11-15 13:34
marcansigh
#asahi-dev · 2021-11-15 13:33
marcanah, linux doesn't listen to the disabled prop on CPUs?
#asahi-dev · 2021-11-15 13:08
marcanI think the fix I pushed should just work then
#asahi-dev · 2021-11-15 13:07
marcanok, so CPU nodes just disappear
#asahi-dev · 2021-11-15 11:42
marcanon the plus side, the extra stuff in t6001 vs t6000 is neatly in a separate PMGR instance, so I can keep my "t6000.dtsi includes t6001.dtsi and removes stuff" thing and all it takes is one node deletion
#asahi-dev · 2021-11-15 11:41
marcanI'm thinking that should at least go in an include :-)
#asahi-dev · 2021-11-15 11:41
marcanin other news, all the pmgr-pwrstate instances for t6001 are ~2000 lines of DT
#asahi-dev · 2021-11-15 11:39
marcanfair enough
#asahi-dev · 2021-11-15 11:18
marcanapparently there is a SetVirtualAddressMap thing, heh
#asahi-dev · 2021-11-15 11:10
marcan(how does EFI do this, I wonder...)
#asahi-dev · 2021-11-15 11:09
marcanbut calling from the kernel into m1n1 has some, er, interesting consequences (e.g. does the kernel 1:1 map m1n1 in the low address space for this, or do the called functions have to be position-independent and callable at any random vaddr the kernel feels like using, or do we turn the MMU off?)
#asahi-dev · 2021-11-15 11:06
marcandoing it the PSCI way should also allow m1n1 to handle the deep cpuidle stuff instead of having to write a driver for that too
#asahi-dev · 2021-11-15 11:03
marcaneither we do an Apple-specific driver, or we add a non-hvc/smc PSCI interface and make m1n1 into a PSCI provider

plus ›