[arch-general] System Freeze on sleep

[Sefa Eyeoglu]

I recently had a similar issue.

As it was breaking my workflow for day-to-day use I investigated the issue.

If the computer is unresposive while resuming it's very likely to be a kernel 
panic.

If there is no display output at all, either the graphics driver itself causes 
the panic, or some kernel module before that.

The only way to debug these kinds of kernel panics was to use a serial port 
and a second computer to get a tty.

As most computers have a COM-header on the mainboard, it's fairly easy to get 
things connected.
I purchased a COM-RS232 PCIe Shield and found a Serial to USB adapter at home.

I then set the following kernel cmdline

	no_console_suspend console=tty0 console=ttyS0,115200

no_console_suspend was needed in my case, but it might not be needed in all 
cases, I am sure it's not recommended to always have that enabled, but as we 
are debugging here we should be fine.

The two console= arguments set the console (kernel log, systemd log and 
panics) to both tty0 (the default) as well as ttyS0 at a baud of 115200 (Check 
if ttyS0 is your serial port while booted).

Now you can get your USB-Serial adapter, connect it with your computer's COM 
port and start a serial console like screen, minicom or picocom.

Now you should see the kernel log while booting. You can even have getty on 
there to login via serial. (I don't know right now if it is autostarted or if 
you need to start it manually with systemctl start getty at ttyS0.service)

Now you can get to debugging: Reproduce your issue and see what you get there.


P.S.
There are other ways to see the console without working graphics. There is 
USB-3.0 debugging (couldn't get that to work myself) and netconsole (didn't 
work for my issue, maybe it wasn't initialised before the panic)

P.S.P.S.
I planned to write this process up somewhere, I guess having it here is a 
start. I may create a Wiki article about this in the future.

P.S.P.S.P.S.
My issue was an actual kernel bug, which I fixed and got merged with recent 
5.11 and 5.12 releases. See https://git.kernel.org/pub/scm/linux/kernel/git/
torvalds/linux.git/commit/?id=7df4ceb60fa9a3c5160cfd5b696657291934a2c9


> Hello All,
> 
> After upgrading to the new kernel I see that my system freezes when
> I try to resume from sleep. This happens intermittently but frequent enough
> to be irritating. I have to reboot the system to recover.
> 
> I am running kernel 5.12.7-arch1-1
> 
> The issue was not seen with kernel 5.11.
> 
> I am running arch on btrfs and I suspect it is a btrfs issue somehow. When I
> check dmesg (journactl --dmesg --boot -1) I see that it always ends with
> the following two lines when the issue is seen.
> 
> May 29 07:12:40 matrix kernel: PM: suspend entry (deep)
> May 29 07:12:40 matrix kernel: Filesystems sync: 0.025 seconds
> 
> I wanted to know how do I attempt to debug this issue?
> 
> Regards,
> Prateek



-- 
Sefa Eyeoglu
https://scrumplex.net 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: This is a digitally signed message part.
URL: <https://lists.archlinux.org/pipermail/arch-general/attachments/20210530/12a568e5/attachment.sig>