#
✕ nick:rqou_
#asahi-dev · 2022-04-07 05:40
rqou_is there somewhere where i can keep track of ongoing work that's too experimental to even make it into the https://github.com/AsahiLinux/linux/tree/asahi/ branch?
#asahi-dev · 2022-04-07 05:36
rqou_just something to note for further investigation
#asahi-dev · 2022-04-07 05:35
rqou_my workaround seems to consistently win the race condition
#asahi-dev · 2022-04-07 05:35
rqou_huh i'm definitely surprised that DCP can break nvme, but i guess it's "fine" for now
#asahi-dev · 2022-04-06 23:46
rqou_anyways, final list of problems: DCP driver doesn't work with sven's iommu patch. the disabling of get_sgtable has to be reverted. after that, a race condition happens with the nvme, which can be worked around by putting DCP into the initramfs
#asahi-dev · 2022-04-06 23:43
rqou_fwiw attempting to not load the DCP driver until the desktop is reached and then manually modprobe-ing it doesn't work either
#asahi-dev · 2022-04-06 23:42
rqou_but yeah, tl;dr everything works
#asahi-dev · 2022-04-06 23:40
rqou_i sure hope i don't end up with major disk corruption due to a hidden bug at some point in the future
#asahi-dev · 2022-04-06 23:40
rqou_looks like i can win the race condition by moving apple_dcp/apple_piodma/appledrm into the initramfs, but keeping the nvme driver linked into the kernel
#asahi-dev · 2022-04-06 23:27
rqou_not exactly the standard thing everyone else is
#asahi-dev · 2022-04-06 23:26
rqou_which is the asahi branch + sven's atcphy + my own m1 max atcphy + DCP + rebased 4k iommu patch
#asahi-dev · 2022-04-06 23:25
rqou_unfortunately i'm running, uh, this kernel https://github.com/rqou/linux/tree/work
#asahi-dev · 2022-04-06 23:25
rqou_perhaps sven could investigate this? https://gist.github.com/rqou/9166943e1d3e102f701beea79ac5e8ea
#asahi-dev · 2022-04-06 23:21
rqou_but DCP itself works
#asahi-dev · 2022-04-06 23:21
rqou_yeah, i seem to be losing the race condition now
#asahi-dev · 2022-04-06 23:11
rqou_very likely there's a race condition?
#asahi-dev · 2022-04-06 23:04
rqou_but setting up an untethered boot doesn't seem to work
#asahi-dev · 2022-04-06 23:04
rqou_no, it was working earlier when i was testing it in the m1n1 hypervisor
#asahi-dev · 2022-04-06 23:01
rqou_but it worked via m1n1 hv
#asahi-dev · 2022-04-06 23:01
rqou_hmm, broke nvme?
#asahi-dev · 2022-04-06 22:55
rqou_DCP works and gets to desktop, with the notch breaking the UI and everything
#asahi-dev · 2022-04-06 22:55
rqou_and it works
#asahi-dev · 2022-04-06 22:55
rqou_yeah, i literally just did
#asahi-dev · 2022-04-06 22:53
rqou_no i care about FEX that much to really want this to work
#asahi-dev · 2022-04-06 22:52
rqou_good job jannau et al
#asahi-dev · 2022-04-06 22:52
rqou_aaand, with that, i have a working DCP driver that boots to the desktop!
#asahi-dev · 2022-04-06 22:51
rqou_https://github.com/jannau/linux/blob/asahi-dcp-nb/drivers/gpu/drm/apple/dcp.c#L469
#asahi-dev · 2022-04-06 22:51
rqou_no it just outright calls that function when mapping buffers for the DCP
#asahi-dev · 2022-04-06 22:47
rqou_but it looks like the new DCP driver indeed requires that function?
#asahi-dev · 2022-04-06 22:47
rqou_because sven apparently wrote "Disable get_sgtable for granule > PAGE_SIZE" "While this function *probably* works correctly without any changes for granule > PAGE_SIZE I don't have any code to actually test it and cannot reason about how the function is supposed to work. Disable it instead until we run into a use case where it's required."
#asahi-dev · 2022-04-06 22:45
rqou_oh huh, this is indeed a 4k kernel limitation
#asahi-dev · 2022-04-06 22:26
rqou_the pile of "denying map of invalid buffer X for pidoma" seems sus, but i don't really know anything about this code or hardware
#asahi-dev · 2022-04-06 22:14
rqou_it still doesn't work, but it gets much further once i remember to actually cherry-pick in the 4k page patch
#asahi-dev · 2022-04-06 22:14
rqou_jannau, j`ey: updated the gist https://gist.github.com/rqou/9015696d4a67a39f1962cf6e6212a06f
#asahi-dev · 2022-04-06 21:56
rqou_i might have accidentally dropped the 4k patch from my kernel
#asahi-dev · 2022-04-06 21:56
rqou_actually, wait
#asahi-dev · 2022-04-06 21:52
rqou_ah, 4k with the patch that's been floating around
#asahi-dev · 2022-04-06 06:57
rqou_oh i see, you're saying it should work with the default "no atcphy module" usb2-only mode that m1n1 sets things up in?
#asahi-dev · 2022-04-06 06:56
rqou_i'll try that later
#asahi-dev · 2022-04-06 06:56
rqou_i saw the trace messages stating that it was switching to USB3 mode
#asahi-dev · 2022-04-06 06:55
rqou_yes, atcphy was enabled, with a cable that only supports usb2
#asahi-dev · 2022-04-06 06:55
rqou_sven: huh, i was specifically testing usb2 and got nothing
#asahi-dev · 2022-04-06 06:40
rqou_fwiw branch dcp _did_ boot until it got to grub, after which grub would complain it was out of memory
#asahi-dev · 2022-04-06 06:39
rqou_jannau: m1n1 branch dcp-nb crashes and reboots on m1 max
#asahi-dev · 2022-04-06 06:30
rqou_does anybody know of a linux tool that can decode a raw prores bitstream? or do i really need to slap some kind of container like mov or mkv around it?
#asahi-dev · 2022-04-06 05:01
rqou_kk, will give that a try. thanks
#asahi-dev · 2022-04-06 03:20
rqou_dunno
#asahi-dev · 2022-04-06 02:24
rqou_iBoot has to hand it to the kernel at some point, but idk if there exists any existing tooling to work with any of this
#asahi-dev · 2022-04-06 02:21
rqou_back in the day there existed ioreg that ran on ios devices, but idk if that works on modern jailbreaks
#asahi-dev · 2022-04-06 02:07
rqou_on a checkm8-able device you should be able to pull what is essentially the ADT info using `ioreg`, but a _lot_ has probably changed between the A6X and the M1

plus ›