<tpw_rules>
what is the timeline like on a 4k kernel? i found another big thing that is broken by 16k pages an am wondering if it's worth getting a patch in nixos...
<chadmed>
im actually building a 4k kernel to test as we speak
<Jamie[m]1>
I'd appreciate a patch in nixos tpw
<Jamie[m]1>
I'd like to run a 16k system indefinitely if possible
<tpw_rules>
Jamie[m]1: well presumably upstream will fix it eventually. it's just that i think this will have to go to staging, and would be a bit of a hack
<tpw_rules>
i am sad because nix just gained the ability to bootstrap on 16k like yesterday...
<tpw_rules>
chadmed: so does this mean someone wrote the necessary patches?
<chadmed>
the DART driver has had 4k support for a while afaik
<chadmed>
not sure if its been tested thoroughly but i plan on changing that over the next few days
<tpw_rules>
ah, ok
<tpw_rules>
let me flip the switch and see what happens
<chadmed>
pretty sure theres nothing else that needs patching to explicitly support 4k pages
<chadmed>
we will find out i guess :)
<chadmed>
ive been in the process of decommissioning my old build machine and havent replaced it yet so im building on my 2017 13" mbp so this is gonna take a while but ill give an update when i get it booted on the j314s
<tpw_rules>
i'lll be done in a couple minutes
<chadmed>
i really shouldnt have offloaded my 5600x so soon :P
<tpw_rules>
it boots but usb doesn't seem to work
<tpw_rules>
i think the problem is IOMMU, not DART
abbas_faiz[m] has joined #asahi-dev
<tpw_rules>
ethernet does not appear to work either, which also needs the IOMMU functioning
<tpw_rules>
so no 4k kernel yet
<tpw_rules>
nvme and the display work fine though
<tpw_rules>
on m1 mini
<chadmed>
should probably add a note on that somewhere in the wiki since 4k pages will be essential for amd64 binary translation
<tpw_rules>
it is being worked on, i don't remember quite by whom
<tpw_rules>
but afaik the intention is to ship asahi linux with a 4k kernel by default
<chadmed>
and even if we dont, distros certainly arent going to be shipping 16k kernels on account of us
chadmed has quit [Remote host closed the connection]
chadmed has joined #asahi-dev
chadmed has quit []
chadmed has joined #asahi-dev
<Glanzmann>
tpw_rules: WIth a 4k kernel you don't have PCIe IIRC. Ethernet and Usb is connected via PCIe IIRC.
<tpw_rules>
yes
<Glanzmann>
At least some of the usb ports, maybe the usb-c work because they don't go through pcie.
<chadmed>
only USB-A goes through PCIe on the M1 Mini
mjg59 has quit [Remote host closed the connection]
the_lanetly_052 has joined #asahi-dev
<jannau>
nothing has changed with wrt 4k support. the darts are configured to bypass mode with 4k page size
<jannau>
that means hardware which which is behind by darts without bypass support (pice/thunderbolt) or locked (dcp) will not work
<jannau>
so display wise only the framebuffer works, okay for now on laptops and imacs, less so on the mac mini especially when using a monitor which is not full hd
<sven>
I have a patch series to actually support 4K page somewhere on the mailing list
<kettenis>
povik: there is another "REG_SERDES_CHANMASK" register at offset 0x18 in the serdes block and if I poke that one I can manage to get data in TDM slot 1 and have "stereo" output on the speaker of the mini
<alyssa>
uh
<kettenis>
uh?
<alyssa>
isn't it only a single speaker
<alyssa>
how does stereo work
<kettenis>
the tas5770 speaker amplifier chip has a mode to mix the left and right channels into a single mono output
<alyssa>
ah, neat
<povik>
kettenis: i know about the mask but i am not sure about its meaning
<povik>
or rather recently i learned that it behaves as if the actual mask was AND of the two masks
<povik>
so to get stereo on mca1, you had to clear a bit in the second mask that is otherwise left set by iboot
<povik>
that's what happened, right?
<povik>
i ran into the same when simulating macbook air's speakers on mini
<povik>
and switching the speaker to be left/right based on the TDM slots it's sensitive to
<kettenis>
yup
<povik>
so in my local tree the MCA driver clears the second mask
<povik>
i am now experimenting with audio capture, there's two masks too, here it may be obvious what they do
<povik>
kettenis: i have m1n1.hw.mca with register definitions
<povik>
i plan to push it together with demo script for capture, once that's ready, but if it would be useful to you i can push it now
the_lanetly_052__ has joined #asahi-dev
chengsun has joined #asahi-dev
<kettenis>
povik: that would probably be helpful
<povik>
on it
the_lanetly_052 has quit [Ping timeout: 480 seconds]
<kettenis>
not all of the register definitions in mca.c on the asahi branch make sense to me
<povik>
kettenis: i am battling git, here it is in paste instead: https://tpaste.us/qP6B
<povik>
but it should essentially be the same as in sound/soc/apple/mca.c, module the RX serdes registers
<povik>
anywhere where there's a cluster SEL it is the cluster no. plus one i think
<kettenis>
with 0 meaning disabled probably
the_lanetly_052__ has quit [Ping timeout: 480 seconds]
Gaspare has joined #asahi-dev
M1bn3mar[m] has joined #asahi-dev
Gaspare has quit [Ping timeout: 480 seconds]
___nick___ has quit [Ping timeout: 480 seconds]
waagrr has quit [Ping timeout: 480 seconds]
waagrr has joined #asahi-dev
chadmed has quit [Read error: Connection reset by peer]
chadmed has joined #asahi-dev
<tpw_rules>
sven: your 4k patches seem to still work on the latest asahi. i had to reorder the includes slightly to get it to apply though
<povik>
i had bit of fun finding which mmio ranges come alive as power domains get enabled one-by-one in a topological order