esv_ has joined #fedora-riscv
esv has quit [Ping timeout: 256 seconds]
zsun has joined #fedora-riscv
zsun has quit [Quit: Leaving.]
davidlt has joined #fedora-riscv
<
davidlt[m]>
So I removed koji-containerbuild for now from our disk images, not really needed :)
<
davidlt[m]>
Fixing other issues right now to get it cooking.
<
davidlt[m]>
I think we might get it this morning, I see: (2435/2435)
<
davidlt[m]>
So DNF went as far as downloading all RPMs. I assume that means all packages are now installable.
<
davidlt[m]>
Yeah, it's installing now.
<
davidlt[m]>
Compression will take few hours so I have time to enjoy on Saturday :D
davidlt has quit [Ping timeout: 256 seconds]
esv__ has joined #fedora-riscv
esv_ has quit [Ping timeout: 256 seconds]
Kevinsadminaccou has quit [Quit: You have been kicked for being idle]
davidlt has joined #fedora-riscv
zsun has joined #fedora-riscv
<
davidlt[m]>
Not a very helpful message.
<
Eighth_Doctor>
rwmjones made that error message
<
Eighth_Doctor>
actually no, someone else, but rwmjones knows about it
<
davidlt[m]>
I am thinking about fuser / lsof / something else to print what is still hanging around using the mount.
<
Eighth_Doctor>
would probably be helpful I guess
<
davidlt[m]>
Because I don't exactly know if whatever is still running is critical or not :/
<
davidlt[m]>
I will experiment today on it.
<
Eighth_Doctor>
cool
davidlt has quit [Ping timeout: 256 seconds]
zsun has quit [Quit: Leaving.]
davidlt has joined #fedora-riscv
<
davidlt[m]>
Conan Kudo: Here are the 1st results:
<
davidlt[m]>
So lsof show "systemd-l" is holding on the tty0
<
davidlt[m]>
Interesting that fuser doesn't show it. I wonder if "sytemd-l" just needs a bit more time to finish.
<
Eighth_Doctor>
probably
<
Eighth_Doctor>
you might be hitting timing issues with how slow RISC-V is
<
davidlt[m]>
I will add some sleep
<
davidlt[m]>
Oh yeah, there are plenty of those. Try running systemd on a single-core VM/machine, or on FPGA :)
<
davidlt[m]>
Things start to break quite often at those configs.
<
davidlt[m]>
I wonder if lsof could show a full command
ahs3[m] has left #fedora-riscv [#fedora-riscv]
ahs3[m] has joined #fedora-riscv
<
davidlt[m]>
So it's +c0 option
<
davidlt[m]>
I wonder what command is systemd-l*
<
davidlt[m]>
Actually lsof is already used imgcreate, but it was never listed as Requires: in the spec.
<
davidlt[m]>
There are a lot of time.sleep(x) too.
<
davidlt[m]>
Ok, 2nd debug patch is now injected.
<
davidlt[m]>
Waits 10 minutes before calling umount (that's plenty of time, and I want the image).
<
davidlt[m]>
Also prints full commands in lsof output, and dump pstree with command, arguments and pids so I can see the whole chain.
<
davidlt[m]>
Realistically it would be enough to use lsof, wait for X minutes until the pids exit.
<
davidlt[m]>
In 1-2 hours we will have more details, I guess.
<
davidlt[m]>
It failed again.
<
davidlt[m]>
As expected this is `systemd-logind`
esv__ has quit [Remote host closed the connection]
<
davidlt[m]>
Conan Kudo: my brain ain't working this late evening, but thinking a bit about it.
<
davidlt[m]>
I assume /sys was bind-mounted into the chroot.
<
davidlt[m]>
The problem is that from the host side systemd-logind holds "active_fd" (/sys/devices/virtual/tty/tty0/active).
<
davidlt[m]>
This is how systemd know which is virtual terminal is active currently.
<
davidlt[m]>
Not all mounts are equal (probably).
<
Eighth_Doctor>
if you want to make a fix, I'd be happy to review/merge it
<
davidlt[m]>
I am thinking that in BindChrootMount if mount point src is /sys , /proc, etc it's safe to lazy unmount
<
davidlt[m]>
I am planning to launch another experiment before I go to sleep
<
davidlt[m]>
ok, it's cooking, and I am off to bed :)
davidlt has quit [Ping timeout: 268 seconds]