marcan changed the topic of #asahi to: Asahi Linux: porting Linux to Apple Silicon macs | Not ready for end users / self contained install yet. Soon. | General project discussion | GitHub: https://alx.sh/g | Wiki: https://alx.sh/w | Topics: #asahi-dev #asahi-re #asahi-gpu #asahi-stream #asahi-offtopic | Keep things on topic | Logs: https://alx.sh/l/asahi
<tpw_rules>
this diff: https://pastebin.com/NzJEQJDW combined with the command line flags '--in-process-gpu --no-sandbox --no-zygote' produces a chromium that works with 16k pages. someone who cares more about google should probably file a bug report
yuyichao has joined #asahi
darkapex1 has joined #asahi
nobodynada has joined #asahi
darkapex has quit [Ping timeout: 480 seconds]
nobodynada has quit [Ping timeout: 480 seconds]
caef^ has quit [Ping timeout: 480 seconds]
caef^ has joined #asahi
skipwich has quit [Ping timeout: 480 seconds]
<marcan>
tpw_rules: should be 64K tbh
<tpw_rules>
yeah it should
<tpw_rules>
that might fix those flags too, that loongson thing is the only thing in there with 16k pages and who knows how maintained it is. but i tried just to try, i'm not actually a chrome user
phiologe has joined #asahi
skipwich has joined #asahi
marvin24_ has joined #asahi
marvin24 has quit [Ping timeout: 480 seconds]
nobodynada has joined #asahi
nobodynada has quit [Ping timeout: 480 seconds]
Dcow_ has joined #asahi
Dcow_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
nobodynada has joined #asahi
Dcow_ has joined #asahi
nobodynada has quit [Ping timeout: 480 seconds]
Dcow has quit [Ping timeout: 480 seconds]
<amw>
FWIW wifi take 3 works perfectly on my mba M1 2020 : nvme rootfs + wifi + touchpad + keyboard all working
<amw>
Makes for a really nice fast light untethered laptop - although I am slightly paranoid about power as there is no way (I know of) to read the battery level...
<Glanzmann>
It will soom with smc marcan told a few days ago. You can also use it for an hour and check in macos for now.
<Glanzmann>
amw: Which Distribution are you running/
<Glanzmann>
?
<amw>
Glanzmann: Bullseye
<amw>
For now I just plug into the charger when I'm not using it, haven't run MacOS for 3 days or so
<Glanzmann>
amw: Nice. Can you upload your .config and xorg tweaks for the touchpad and paket list soemwhere?
<Glanzmann>
Basically you create the partition from macos and dd the image using the tpw_rules 'tar as gzcat' trick.
<Glanzmann>
Than you can resize2fs to fill the partition from the booted system.
<amw>
ok - sounds nice. I'm too scared to move off my 11.4 macOS at the moment - after seeing your struggles ...
<amw>
Hope it's working for you again
<Glanzmann>
amw: For the notebooks it is fine. The mini is the problem.
<amw>
That's good to hear.
<Glanzmann>
I tried to reinstall on the mini in order to do a bullseye quickstart installation video using a hdmi grabber ...
<Glanzmann>
But to be honest I hope that one of the best, maybe marcan, initializes the hdmi in m1n1 maybe also make it possible to set a resolution and than mini is usable again.
par0 has joined #asahi
<Glanzmann>
s/best/devs/
par0 has quit []
<Glanzmann>
amw: Because it happened to me once, do not close the lid when Linux is running, it gets warm under there since we have no ability to disable the screen at the moment.
<amw>
ok I did that - it didn't seem too hot, a little warm under the hinge area but fine otherwise
<marcan>
pushed wifi/take4 which unbreaks t6000 machines (which broke due to the T2 changes)
<Glanzmann>
amw: I left it sitting like that for several hours. When I cam back the metal besides the touchpad was very warm. I never felt it that warm, but it cooled down in a minute or so.
<Glanzmann>
amw: Thanks a lot might be helpful if wifi still does not work out of the box for me, thanks a lot.
<Glanzmann>
amw: Did you do anything special for the touchpad? Are you using xorg or wayland?
MajorBiscuit has joined #asahi
<amw>
Glanzmann: Sorry I was busy - I believe I am just running X11 : /usr/lib/xorg/Xorg fluxbox, firefox runs very well (no sound of course)
<amw>
The only problem I had was with the touchpad was fixed when I removed the xserver-xorg-input-synaptics and installed the xserver-xorg-input-libinput
<amw>
I also have the xserver-xorg-input-mtrack installed - not sure if it is needed
Dcow has joined #asahi
Dcow__ has joined #asahi
Dcow_ has quit [Ping timeout: 480 seconds]
arahael3 has quit [Quit: WeeChat 3.1]
Dcow_ has joined #asahi
nobodynada has joined #asahi
Dcow has quit [Ping timeout: 480 seconds]
Dcow__ has quit [Ping timeout: 480 seconds]
Major_Biscuit has joined #asahi
MajorBiscuit has quit [Ping timeout: 480 seconds]
ChaosPrincess has quit [Quit: WeeChat 3.3]
ChaosPrincess has joined #asahi
pg12_ has joined #asahi
pg12 has quit [Ping timeout: 480 seconds]
<mps>
amw: for me keyboard/touchpad works with both, synaptics and libinput driver, just had to configure which one to use
<landscape15[m]>
Is it possible to install Linux on a FAT32 partition?
<landscape15[m]>
I created it in macOS
<mps>
landscape15[m]: maybe but it will not work correctly because FAT can't preserve file names and rights
<VinDuv>
it may work but probably not very well
<Redecorating[m]>
also no symlinks
<VinDuv>
non-journaled case-sensitive HFS+ may work though
<landscape15[m]>
VinDuv: why? Is there a driver?
<mps>
marcan: just tested with wifi/take4 branch on MBP (j293) and wifi still doesn't work, same error in dmesg out: brcmfmac 0000:01:00.0: brcmf_pcie_init_ringbuffers: Allocating ring buffers failed
<mps>
I know latest changes are to something else but wanted to inform you
<landscape15[m]>
mps: btw i will format the partition directly in Linux, booted from USB (and then `cp` to the new partition).
<mps>
landscape15[m]: linux distros usually have a lot of symlinks but FAT doesn't supports that
<mps>
you are free to do whatever you want but advice is: don't waste time
<landscape15[m]>
I mean I will format the partition in ext4 (not FAT) from Linux
<mps>
landscape15[m]: ah, ok
<mps>
use 'cp -a ....' to preserve attributes and time on files
<amw>
mps: Didn
<amw>
mps: Didn't mess with the config for synaptics - don't really follow all of the features
<amw>
landscape15[m]: Definitely agree unless you are looking for adventure stick with ext4 as I had wierd errors like ping for normal users doesn't work because I was missing the ext4 attributes feature missing
chadmed has joined #asahi
<mps>
amw: https://tpaste.us/ZjOo this is simple config which uses synaptics for touchpad and libinput for keyboard
<amw>
mps: I'll try it and see if makes a difference
<as400[m]>
tpw_rules: sandboxing is not working in chromium due to missing kernel bits
<as400[m]>
tpw_rules: CONFIG_MODULE_ALLOW_MISSING_NAMESPACE_IMPORTS=y and CONFIG_SECURITY_YAMA=y. First one just to stay on the safe side - I'm not sure it's needed.
aleasto has joined #asahi
<amw>
mps: Yep - I install the debian bullseye synaptics package and touch screen didn't work - but mouse did - with this config https://paste.debian.net/1224616/
<amw>
I switched it for your configuration and the touchpad worked but my USB mouse didn't...
<mps>
amw: touch screen? maybe you mean touchpad?
<marcan>
mps: can you enable all the debug options and paste a full dmesg of at least the brcmfmac range? CONFIG_BRCM_TRACING=y CONFIG_BRCMDBG=y
<mps>
amw: you have to add section for mouse if you use it, or any other device
<marcan>
though I probably won't work on this until after xmas :p
<amw>
Sorry - yep touchpad (mba) - this one in xserver-xorg-input-libinput allows both mouse and touchpad for me - Sorry I mixed them up - this is https://paste.debian.net/1224617 is the synaptics config
<mps>
marcan: yes I can, is these two enough to enable
<j`ey>
marcan: no xmas stream? :( :P
<marcan>
unless you want to see me eating cheese fondue with a friend and opening presents :p
<marcan>
mps: yes
<mps>
ok
<j`ey>
marcan: that sounds great, enjoy!
nobodynada has quit [Ping timeout: 480 seconds]
<mps>
hm, not enough just these two, have to enable more codedump and some other. ok will build complete new kernel and reboot
Major_Biscuit has quit [Ping timeout: 480 seconds]
<jannau>
mps: not sure if the debug message is wrong or if it loads indeed firmware brcm/brcmfmac4378b1-pcie.bin instead of brcm/brcmfmac4378b1-pcie.apple,honshu.bin
<mps>
jannau: there is no brcmfmac4378b1-pcie.bin in /lib/firmware/brcm
<mps>
would someone who extracted firmware with latest installer post tarball to me. maybe my extraction isn't proper
<jannau>
mps: please load brcmfmac with 'debug=0x00080006'
<jannau>
mps: most obvious difference to j274/j314c is "brcmf_pcie_init_ringbuffers Using host memory indices"
<jannau>
that's the else part of your 'brcmf_pcie_init_ringbuffers Using TCM indices'
<jannau>
I'd say it's doubtful that the macbook pro 13 should use a different method
___nick___ has joined #asahi
___nick___ has quit []
___nick___ has joined #asahi
<mps>
my C skill faded away to hack this
<mps>
I see code but don't understand it
<jannau>
are you running a kernel with 4k or 16k page size?
<mps>
4K
<jannau>
I think the pcie dart don't suppport bypass and don't work currently with 4k kernel page size
<jannau>
please try with a 16k kernel
<mps>
jannau: oh
<mps>
and f2fs works only with 4K
<mps>
ok will test
<jannau>
I think the dma_alloc_coherent in brcmf_pcie_init_ringbuffers() fails and the driver tries to use "TCM indices" instead. no idea why the latter fails or if it is supposed to work
<mps>
iirc on one other intel machine brcmfmac works with 4K pages, but have to check to be sure
nobodynada has joined #asahi
<jannau>
this is unrelated to brcmfmac. the iommu for the pcie ports works currently just with 16k kernel page size
chadmed has quit [Remote host closed the connection]
<mps>
finished build, now will reboot and see
<mps>
jannau: it works with 16K pages
<mps>
now I have to convert my disks from f2fs to something else, ufff
<mps>
jannau: thanks for helping find cause of problem
<mps>
only left 'apple-pmgr-pwrstate 23b700000.power-management:power-controller@2c0: PS mca1: Failed to reach power state 0xf (now: 0x24f)' this to not have usb dongle attached all day
nobodynada has quit [Ping timeout: 480 seconds]
Major_Biscuit has joined #asahi
yuyichao has quit [Ping timeout: 480 seconds]
Major_Biscuit has quit [Ping timeout: 480 seconds]
nobodynada has joined #asahi
caef^ has quit [Ping timeout: 480 seconds]
<Glanzmann>
mps: I did not know that you run a 4 kB kernel, otherwise I could have told you as well. :-( I heard had I think sven wanted to make the dart work 4k soon ...
<Glanzmann>
mps: Otherwise ext4fs give me good results. :-)
<Glanzmann>
jannau: Good catch with mps wifi.
<mps>
Glanzmann: also I remember that sven told something about dart on 4K
<mps>
Glanzmann: ext4 on flash devices is slow with journal and not safe without journal
<mps>
but for now will switch to it
<mps>
(better slow than not working)
jochum_ has joined #asahi
jochum has quit [Ping timeout: 480 seconds]
nobodynada has quit [Ping timeout: 480 seconds]
yuyichao has joined #asahi
<tpw_rules>
as400[m]: just curious, did you actually try to build and run it on a 16k kernel? the symptoms i saw were that it crashed instantly on startup without those flags
<tpw_rules>
with a whole bunch of log messages that didn't really make sense
<tpw_rules>
(mostly seemed to be the crash reporting process crashing)
<Glanzmann>
mps: I have ext4 on a lot of flash devices, but had never any issues. Journaling is on by default, isn't it?
<Glanzmann>
For the m1, I often prefix apt and kernel builds with 'eatmydata' to delay sync.
<Glanzmann>
Marry chistmas to everyone.
<mps>
Glanzmann: yes, journaling is on if you don't disable it
<mps>
and it 'lead' to wearing flash devices
<mps>
f2fs is specifically made for flash devices
<mps>
also, Marry Christmas to all
<mps>
Glanzmann: iwd works nice with wifi now
<mps>
don't know if debian have iwgtk packaged, simple gui for iwd
<mps>
and iwd can work in STA, AP and Ad-Hoc mode
<Glanzmann>
mps: To be honest, I do it the following with debain. I have this giant wpa_supplicant.conf with all the wifi access points I used in the past, all over the world. And than I have this twoliner: https://pbot.rmdir.de/X2A0alscskfEEp4uCHK_ow and the rest normally happens automatically ...
<mps>
well, I use wpa_s only with dongles which have wext interface. iwd is lot faster on net setup because it has FILS implemented
<Glanzmann>
mps: I see. I had no issues with failing flash devices the last 10 years until this year. This year 15 Samsung SSDs (EVO, QVO, NVMe Pro) failed with less than 10 TB written. With EVO and QVO there is a report about them failing with ncq trim enabled, so I disabled it, and another one failed. So for now I'm going with Intel DC edition and Micron DC edition SSDs. It also seems that the enterprise version
<Glanzmann>
of samsung nvmes is also working properly.
<Glanzmann>
mps: I see, never heard of iwd, need to try it out. And honestly these days I'm only using a single ssid and when I travel I get a sim card and use wifi tethering with the very same ssid ...
<Glanzmann>
mps: I use autologin on my Linux systems, and when I'm logged in, wifi is also there even with wpa_supplicant.
<mps>
Glanzmann: I had to add delay to iwd on boot because it was too fast on two chromebooks :)
the_lanetly_052 has joined #asahi
nobodynada has joined #asahi
amw has quit [Remote host closed the connection]
amw has joined #asahi
skoobasteeve_ has joined #asahi
skoobasteeve has quit [Ping timeout: 480 seconds]
<as400[m]>
tpw_rules: I didn't build chromium with your patch. But I have seen chromium not starting without kernel options I mentioned. So, to make things clear, options mentioned are required to start the browser with sandbox mode.
nobodynada has quit [Ping timeout: 480 seconds]
<ey3ball[m]>
Had some issues with my rootfs getting trashed ... looks like MacOS can corrupt partitions if it think it knows the partition type set on the GPT
<ey3ball[m]>
s/Had some issues with my rootfs getting trashed ... looks like MacOS can corrupt partitions if it think it knows the partition type set on the GPT/Had some issues with my rootfs getting trashed ... looks like MacOS can corrupt partitions if it thinks it knows the partition type set on the GPT .../
<tpw_rules>
ey3ball[m]: that's a little worrying. were they set as FAT?
<marcan>
mps: wait, is this a 4K kernel or a 16K kernel?
<marcan>
4K kernels are not supported and will not work for PCIe devices yet, until the patchset for that gets in shape
<ey3ball[m]>
Yep type was set to 0700, no more issue, after updating the partition type to 8300
<marcan>
that would explain that kind of error
<marcan>
I don't see how that error can happen other than if DMA allocs are horribly broken, which they would be on a 4K kernel
<jannau>
marcan: yes, mps was using a 4k kernel. wlan works after recompiling with 16k page size
<marcan>
oh, just saw, heh
<mps>
marcan: yes, it was 4K pages kernel
<mps>
but lscpi shows even on it devices, only ringbuffer couldn't be allocated
<marcan>
yes, DMA doesn't work
<marcan>
that is expected
<marcan>
devices will show up but the drivers will all error out
<mps>
is that hardware limitation
<marcan>
yes, the IOMMUs are 16K only
<marcan>
these machines are not designed for running with 4K pages in kernelmode
<mps>
long time passed when I worked directly with low level devices
<marcan>
sven came up with some heroic patches to work around this and allow Linux to use 16K pages for IOMMUs while using 4K pages for the CPU, but they aren't ready yet
<mps>
ah, now understand
<mps>
(my first serious IT job was to design industrial computer 30 years ago)
<marcan>
I don't think IOMMUs existed back then :p
<mps>
right, but MMU existed
<marcan>
yes
<mps>
but about 20 years ago I had to switch to work on financial and DB field, and previous knowledge faded away mostly
<kettenis>
marcan: the sun4m architecture included an IOMMU!
<marcan>
okay, okay, 33 years ago, you win :p
<mps>
I'm not sure if the Z8000 family had IOMMU in some of devices
<mps>
but PMMU was there for sure
<mps>
anyway pmgr issue I reported is not iommu related I guess :)
daniel0611[m] has joined #asahi
caef^ has joined #asahi
<marcan>
pmgr issue is just due to a hack in the MCA driver
<marcan>
or rather missing feature
<mps>
aha, so iiuc then povik 'have' to fix it
<mps>
what about pcie enable device python script, should it be added in m1n1 or kernel
<mps>
(to get rid of the cables and another machine on the desk)
nobodynada has joined #asahi
martinstewart has joined #asahi
<martinstewart>
hello
<martinstewart>
hows everyones days
nobodynada has quit [Ping timeout: 480 seconds]
<martinstewart>
now i tihnk of it i should probaly be in asahi offtopic to say hi to everyone eh im shure marcan wont mind
<martinstewart>
im shure i would of been toled off by now if it wasent allowed
<mps>
martinstewart: hi, saying hello is not problem imo, but I'm not OP of channel
Gaspare has joined #asahi
<jn>
martinstewart: saying hi is ok, but people might not respond to it
martinstewart has quit [Ping timeout: 480 seconds]
Major_Biscuit has joined #asahi
Major_Biscuit has quit []
joske has joined #asahi
MajorBiscuit has joined #asahi
<j`ey>
mps: thats being added to the kernel
martinstewart has joined #asahi
nobodynada has joined #asahi
<mps>
j`ey: which branch
<j`ey>
mps: n/a yet
<mps>
ah
<mps>
j`ey: ok, will wait new asahi branch
<Glanzmann>
mps: About the pci expereiments features, if I understood correctly, than we get SMC support soon and than that will enable the pci devices.
yuyichao has quit [Ping timeout: 480 seconds]
<mps>
Glanzmann: I think it is System Management Controler and not sure enabling PCI depends on it, but that is my guess only
<tpw_rules>
because the SMC has to turn on the power to the wifi card
<NightRaven[m]>
any news on getting wifi dongle working on the macbook pro ?
nobodynada has quit [Ping timeout: 480 seconds]
<mps>
CONFIG_ARM_SCMI_PROTOCOL
<tpw_rules>
not sure that's relevant
<tpw_rules>
apple has always had their own SMC
<NightRaven[m]>
* on getting a wifi dongle
<mps>
tpw_rules: yes, but iirc CONFIG_ARM_SCMI_PROTOCOL is top one and other specific for SOC are 'down' of it
<mps>
NightRaven[m]: soon will be asahi branch refreshed with internal wifi working, I hope
<NightRaven[m]>
mps: oh ok can't wait : )
<NightRaven[m]>
🤞
<mps>
NightRaven[m]: also I ;)
<mps>
NightRaven[m]: if you have another machine you run it right now by building kernel from wifi/take4 branch
<mps>
you can*
<NightRaven[m]>
mps: sadly i don't have another machine : (
<mps>
NightRaven[m]: sorry to hear this :-(
<NightRaven[m]>
wait
<NightRaven[m]>
i just asked my friend if i can borrow his laptop
<NightRaven[m]>
and he said yesss
<NightRaven[m]>
lets go
<NightRaven[m]>
but it has windows
<mps>
uh, I wouldn't try this
<tpw_rules>
yeah that's not a good idea
<NightRaven[m]>
i can boot linux from a usn
<NightRaven[m]>
s/usn/usb/
<NightRaven[m]>
and try from their
<NightRaven[m]>
tpw_rules: but why ?
<tpw_rules>
because you won't even be able to boot it without using their computer
<mps>
you will have to install m1n1 from github and a lot of other pkgs
<tpw_rules>
and building a kernel is not exactly a quick 5 minute thing, in most cases
<NightRaven[m]>
oh so i need another macbook pro
<NightRaven[m]>
?
<NightRaven[m]>
macbook*
<mps>
NightRaven[m]: no
<tpw_rules>
no, you just need a computer you own that runs linux with an x86_64 or so cpu
<NightRaven[m]>
i can install linux alongside windows
<mps>
any linux/bsd computer with good usb is enough
<NightRaven[m]>
but i have to ask permission
<NightRaven[m]>
mps: ohh ok
<NightRaven[m]>
i already have a linux mint usb stick
<mps>
and you need proper usb-c cable
<NightRaven[m]>
mps: i have that too
<mps>
not all usb-c cables work
<mps>
I have two which are problematic
<NightRaven[m]>
like i have an apple usb cable
<NightRaven[m]>
but i have a usb c to usb 3 adapter
<mps>
and, what's worse, with bad cables interfaces could be damaged
<NightRaven[m]>
but can i still try it ?
<NightRaven[m]>
since i have 2
<mps>
NightRaven[m]: ofc you can and I don't want to 'block' you