#
✕ nick:povik
#asahi-dev · 2023-02-23 19:26
poviki did the same m1n1's display.c is doing to reallocate framebuffer
#asahi-dev · 2023-02-23 19:26
povikmaybe i am wrong in removing the allocated regions from the usable memory range
#asahi-dev · 2023-02-23 19:25
povikit must have something to do with them being placed at the top of memory and allocated in m1n1, because the firmware regions passed down from apple's carveouts don't trigger that
#asahi-dev · 2023-02-23 19:20
poviku-boot not happy about the sio reservations ^--
#asahi-dev · 2023-02-23 19:20
povikERROR: reserving fdt memory region failed (addr=9e09d4000 size=4000 flags=4)
#asahi-dev · 2023-02-23 19:20
povikERROR: reserving fdt memory region failed (addr=9e09b4000 size=20000 flags=4)
#asahi-dev · 2023-02-23 19:20
povikERROR: reserving fdt memory region failed (addr=9e09b0000 size=4000 flags=4)
#asahi-dev · 2023-02-23 19:20
povikERROR: reserving fdt memory region failed (addr=9e09ac000 size=4000 flags=4)
#asahi-dev · 2023-02-23 19:20
povikof course i haven't set it up properly
#asahi-dev · 2023-02-23 18:41
povikso why don't i get the early proxy here :p
#asahi-dev · 2023-02-23 18:41
povikok
#asahi-dev · 2023-02-23 18:40
povikah, so that's not the other thing after &&?
#asahi-dev · 2023-02-23 18:35
povikhttps://github.com/AsahiLinux/m1n1/blob/main/src/main.c#L78
#asahi-dev · 2023-02-23 18:35
povikwhat is the purpose of the !display check here?
#asahi-dev · 2023-02-23 18:34
povik--
#asahi-dev · 2023-02-23 18:17
poviknot that i know what those are, or where they are at
#asahi-dev · 2023-02-23 18:17
povikit might make sense for us to push the kernel cap even lower due to other-than-thermal limits
#asahi-dev · 2023-02-23 18:16
povik(i have a duct-tape admac stream tracer somewhere with which to check that)
#asahi-dev · 2023-02-23 18:09
poviknot sure how much significance we should attach to the 3dB pad since we don't know how much of the codec input range is exercised by apple's EQ
#asahi-dev · 2023-02-23 18:08
povikmarcan: cool!
#asahi-dev · 2023-02-19 14:37
povikanyway now it's clear
#asahi-dev · 2023-02-19 14:37
povikah, i wasn't looking at the latest version
#asahi-dev · 2023-02-19 14:33
povikright, now it makes sense
#asahi-dev · 2023-02-19 14:33
povikwith the example being a 1:1 mapping of a region
#asahi-dev · 2023-02-19 14:32
povikaha, the upstream series merely reserves the iovas, doesn't set up translation for those but assumes some other mechanism resolves those addresses
#asahi-dev · 2023-02-19 14:19
poviksetting up translation from iova space seems to be the whole point of the series :D
#asahi-dev · 2023-02-19 14:19
povikwhat you mean by translation here?
#asahi-dev · 2023-02-19 14:07
povikif you haven't seen there's this standard way to arrange the mappings ^-- (not sure if that was the final version of the series)
#asahi-dev · 2023-02-19 14:06
povikhttps://lore.kernel.org/all/20220905170833.396892-1-thierry.reding@gmail.com/
#asahi-dev · 2023-02-19 13:44
poviksven: yeah, that's exactly right
#asahi-dev · 2023-02-19 11:07
povikthat's the 128mb i was referring to earlier ;)
#asahi-dev · 2023-02-19 11:05
poviki don't see it
#asahi-dev · 2023-02-19 11:04
povikchadmed: what do you mean by the default heap size in hv/__init__?
#asahi-dev · 2023-02-19 11:02
povikif you find the root cause it would be good to fix
#asahi-dev · 2023-02-19 11:02
povikstill it's a bug it silently overrides/crops the data instead of erroring out
#asahi-dev · 2023-02-19 09:44
povikyeah
#asahi-dev · 2023-02-19 09:42
povikunless there's some trailer it's checking for
#asahi-dev · 2023-02-19 09:42
poviksanity check probably wouldn't see that the data past a point are missing
#asahi-dev · 2023-02-19 07:54
povikand compressed_writemem uses the heap internally
#asahi-dev · 2023-02-19 07:54
poviki am surprised this doesn't error out on that large a payload since run_guest.py allocated 128mb for the heap
#asahi-dev · 2023-02-19 07:53
povikhttps://github.com/AsahiLinux/m1n1/blob/v1.2.4/proxyclient/m1n1/hv/__init__.py#L1658
#asahi-dev · 2023-02-19 07:51
povikisn't it simply a case of the hypervisor cropping the payload or otherwise damaging it past a certain size limit?
#asahi-dev · 2023-02-17 13:49
povikwhere the consequences are more dire
#asahi-dev · 2023-02-17 13:49
povikand i only found out because i copied the relevant code into the sio driver
#asahi-dev · 2023-02-17 13:49
povikhuh, upstream admac has a memory leak
#asahi-dev · 2023-02-14 21:11
povik(playback uses 32-bit slots though)
#asahi-dev · 2023-02-14 21:11
poviksure, but there are no 16-bit floats nor 32-bit TDM slots in the picture, so that comment was just misleading
#asahi-dev · 2023-02-14 20:39
povikgive your comments on the sio binding or live with it for eternity!
#asahi-dev · 2023-02-14 12:59
povik:)
#asahi-dev · 2023-02-14 12:58
povikaudiowise nothing but the bits branch is relevant

plus ›