jaeger changed the topic of #crux to: CRUX 3.7 | Homepage: https://crux.nu/ | Ports: https://crux.nu/portdb/ https://crux.ninja/portdb/ | Logs: https://libera.irclog.whitequark.org/crux/
_0bitcount has quit [Quit: Leaving]
tilman has quit [Ping timeout: 245 seconds]
tilman has joined #crux
lavaball has quit [Remote host closed the connection]
ppetrov^ has quit [Quit: Leaving]
mnkydeth has quit [Read error: Connection reset by peer]
woland has joined #crux
woland has quit [Client Quit]
woland has joined #crux
woland has quit [Remote host closed the connection]
zorz has joined #crux
lavaball has joined #crux
nekobit has quit [Ping timeout: 245 seconds]
SiFuh has quit [Remote host closed the connection]
SiFuh has joined #crux
<cruxbot> [contrib.git/3.7]: amdgpu_top: 0.5.0 -> 0.6.0
<cruxbot> [contrib.git/3.7]: fzf: 0.44.1 -> 0.45.0
<cruxbot> [contrib.git/3.7]: python3-frozendict: 2.3.10 -> 2.4.0
<cruxbot> [contrib.git/3.7]: python3-netaddr: 0.9.0 -> 0.10.0
tarxvfz has joined #crux
nekobit has joined #crux
<SiFuh> Anyone running an AMD Ryzen with the 6.1.X kernel ?
<ukky> SiFuh: I suspect that having external USB root device makes udev print '^@'. Not sure if it fixes this issue, but did you add 'rootwait'. It is required at boot, but maybe helps in re-enumerating devices too when udev starts.
<SiFuh> ukky: I'll have a look later. Building the the 6.6.9 kernel at the moment to see if it has anything useful.
<SiFuh> I don't think rootwait will do much since the system is already running. But I could try just for the thrills
<jaeger> I run a couple of ryzen systems with 6.1.x
<jaeger> No ^@ on them
<SiFuh> Cheers
<SiFuh> jaeger: What processor family do you choose in your kernel configuration?
<jaeger> Opteron/Athlon64/Hammer/K8
<SiFuh> Thought so. I was reading it in the Gentoo forums
<SiFuh> jaeger: "It is possible to incorporate the microcode firmware blob files for all AMD processors in the Linux kernel".
<SiFuh> All at the same time. Heh
<jaeger> Indeed
<SiFuh> I never knew that.
<jaeger> I remember reading it somewhere but haven't tried to :)
<SiFuh> Probably here
<jaeger> Could be, yeah
_0bitcount has joined #crux
<cruxbot> [contrib.git/3.7]: csvlens: 0.4.0 -> 0.5.0
<cruxbot> [contrib.git/3.7]: qownnotes: 24.1.0 -> 24.1.1
tarxvfz has quit [Ping timeout: 240 seconds]
tarxvfz has joined #crux
lavaball has quit [Remote host closed the connection]
lavaball has joined #crux
<ukky> SiFuh: Try running as root: http://0x0.st/H6oG.sh
<SiFuh> ukky: For every device. Trigger ADD event for device ...... unable to open the device .......
<ukky> Is '/sbin/udevd --daemon' running?
<SiFuh> Yep
ivandi has quit [Quit: WeeChat 4.1.2]
<ukky> What is the output?: mount | grep ' on /\(dev\|sys\) '
ivandi has joined #crux
<ukky> unrelated to ^@, why do you mount /sys twice?
<SiFuh> I don't know where this one comes from. No other machine has it. I am using the latest iso from jaeger's updates sysfs on /sys type sysfs (rw,nosuid,nodev,noexec,relatime)
<ukky> One mount location for /sys is at /etc/rc
<SiFuh> Yes the bottom one
<ukky> Exacly, as the 'device' parameter is set to 'none'
<SiFuh> I plan to reboot soon. I am just finishing the touches on the 6.6.9 kernel. There is an issue with F2FS on a root partition after 6.2.x and it seems to remain even until 6.6.x. I will then check the mount |grep again when it is ready
ivandi has quit [Quit: WeeChat 4.1.2]
<SiFuh> ukky: Whatever it is, it happens on boot.
<SiFuh> I disabled xdm and ntpd. So rc.conf is only loading what CRUX defaults to. The only two things I can think of is dracut and the kernel
<ukky> Probably dracut then. I guess it is initramfs tool, right?
<SiFuh> yes
<SiFuh> I have another CRUX box using dracut, I will compare to that.
<ukky> You can add 'mountpoint -q /sys && umount /sys' before '/bin/mount -t sysfs none /sys' in /etc/rc to fix /sys mounted twice
<SiFuh> jaeger: and farkuhar: for the record if anyone has trouble booting F2FS root for kernel versions 6.2.x and greater (Currently tested to 6.6.9) they will need to append 'rootflags=noflush_merge' to the kernel command line
<SiFuh> ukky: If it is a dracut issue, then it might be something that needs the CRUX dev's to modify since a few of the users do use dracut
<ukky> Mounting /sys twice does not affect overall functionality. It is just wrong to do that, as initial mount will be hidden by the latest mount. Ideally, this should be fixed.
<SiFuh> Also re-ran your script on a fresh boot and still just shows those errors
<ukky> Do you still run start_udev at boot? Original start_udev or modified?
<SiFuh> Original
<ukky> Please upload /tmp/trace.txt from this (modify path to script): doas sh -c 'sh -x /tmp/trigger_device_add.sh' >/tmp/trace.txt 2>&1
<SiFuh> ix.io is gone... damn it
ivandi has joined #crux
<ukky> SiFuh: What does this print?: ls -ld /sys/devices/cpu
<SiFuh> drwxr-xr-x 6 root root 0 Jan 2 03:00 /sys/devices/cpu
<ukky> correct
ivandi has quit [Quit: WeeChat 4.1.2]
<SiFuh> ukky: On all my systems running dracut, the same results for the double mounted /sys
ivandi has joined #crux
<ukky> My /etc/runit/1 has this line: mountpoint -q /sys || mount -o nosuid,noexec,nodev -t sysfs sys /sys
<SiFuh> So if it isn't already mounted, it will mount it
<ukky> Exactly
<SiFuh> (rw,nosuid,nodev,noexec,relatime)
lavaball has quit [Quit: lavaball]
<SiFuh> Who's the one that wrote the rc script?
<SiFuh> Might be good to put a few sanity/failsafe checks in
<cruxbot> [opt.git/3.7]: fmt: 10.1.1 -> 10.2.0
<ukky> SiFuh: Please try this script instead: http://0x0.st/H6Hp.sh
<SiFuh> Could not find noisy device
<ukky> What about this command?: doas /sbin/udevadm trigger --type=devices --action=add
<SiFuh> Also silent, I tried that before
<ukky> Modify start_udev: Comment-out '/sbin/udevadm trigger --type=devices --action=add' and add my script '/PATH/script.sh' instead, then reboot
tarxvfz has quit [Remote host closed the connection]
<SiFuh> I did and was silent
<ukky> Modify '/etc/rc' and execute /sbin/start_udev this way: '/bin/sh -x /sbin/start_udev', restore start_udev to original, then reboot
<SiFuh> I inserted the '/PATH/script.sh' in and still silent. When I said I did and was silent, I did it manually.
woland has joined #crux
woland has quit [Client Quit]
<SiFuh> ukky: https://s3.amazonaws.com/i.snag.gy/IWcHdb.jpg Just ignore the snd_hda stuff
<ukky> Another idea: This could be output of '/sbin/udevd --daemon', as it does fork in the background and might generate output at any time
<SiFuh> If I comment out the second last line the ^@ never appears
<SiFuh> Anyway, thanks for trying. I will look into it at a later date. It might be something as simple as a glitch for booting a system from a USB external hardisk
<ukky> I tried, but failed.
<SiFuh> I don't think there was failure. We learned dracut mounts its own /sys and that rc mounts on top of it a second time.
<SiFuh> That to me is more important than a ^@ :-)
<SiFuh> ukky: I put this in the rc script by the way ountpoint -q /sys || /bin/mount -t sysfs none /sys
<SiFuh> Hmm where did the 'm' go?
<SiFuh> ukky: Also, I found out that dracut also mounts /proc as well.
<SiFuh> Seems dracut also mounts /dev/shm as well
<ukky> Do you have two mounts as well for /proc?: mount | grep ' on /proc '
<SiFuh> Yes and shm
<SiFuh> ukky: I did this https://dpaste.com/4VA2Y5TUR.txt
<ukky> My init script does not mount these (if already mounted): /proc /sys /run
<SiFuh> And I commented out the line shm in fstab. I will look deeper into what dracut does later on
<SiFuh> jaeger: ^
<ukky> And these are unmounted (if already mounted), then re-mounted: /dev/pts, /dev/shm, /dev/mqueue, /dev; Because my script 'supports' both udev and mdev
<ukky> SiFuh: Try also this way of starting udev daemon: /sbin/start-stop-daemon --start --quiet --name udevd --pidfile /run/udevd.pid --startas /sbin/udevd -- --daemon
disapper3nce has quit [Quit: leaving]
disapper3nce has joined #crux
disapper3nce has quit [Client Quit]
disapper3nce has joined #crux
<SiFuh> ukky: Later, the machine needs to work now. However, I was poking around dracut and it does load those https://github.com/dracutdevs/dracut/blob/master/modules.d/99base/init.sh
<ukky> SiFuh: Thanks for the link, I will use the same way as dracut to test whether /proc or /sys is mounted, as 'mountpoint' cannot be used before /proc and /sys are mounted
<r0ni> i just used dracut on arm64, didn't think it would work, but it did and i've scared for no reason lol
<SiFuh> r0ni: Do you have double mounts?
<r0ni> it appears so
<r0ni> proc on /proc and none on /proc
<SiFuh> Do you have fusectl as well?
<r0ni> nope, no fusectl
<r0ni> /sys appears to be mounted like /proc tho
<SiFuh> Cheers, I didn't look into fusectl yet. Just noticed the dracut machines have it double mounted
<SiFuh> In my paste, the first line for each is not CRUX default, the second line is. Except fusectl
<r0ni> hrm, interesting. ya I'd just used dracut for first time, after building a new kernel based off gentoo config, so I didn't know if it'd even work, but it booted right up
<r0ni> better than the debian config i'd been using at least
<SiFuh> It is really easy
<SiFuh> r0ni: https://dpaste.com/2X43WPTTM.txt Config setup from vccrux which I am using for now.
<r0ni> i was doubtful dracut was setup to work on arm64, sure it easy, but nothing ever works on arm
<SiFuh> Haha
<r0ni> using that the initramfs.img is much much smaller
<SiFuh> When I have time, I will dissect it later, see what else I can strip out of it
<r0ni> now to see if it works lol
<SiFuh> Should, he mostly removed the systemd junk and added the crypto setup for encrypted disks
<r0ni> i still don't fully trust dracut, so each creation that works... i'm surprised
lavaball has joined #crux
lavaball has quit [Client Quit]
_0bitcount has quit [Remote host closed the connection]