Suspend to RAM broken in Mageia 6

[artee]

Hi all,
my upgrade from Mageia 5 to 6 was a bit of a mixed experience, mostly due to the differences between Plasma in 6 vs KDE in 5.

But the most painful regression is the non-functional suspend to RAM.

Trouble is, I have not had to mess with Linux, under the hood, since way before the systemd days.

The symptoms:
either by menu - logout/power session - suspend (using powerdevil?), or via cli/bash:
systemctl suspend
or
pm-suspend

Result: pc shuts down, after starting it again, it comes up to the final screen before shutting down, typically black with just the mouse pointer, stays there frozen.
Most of the times I tried, it went like that.
Once with pm-suspend, it came back to life completely.
A few times, when it's frozen, the caps-lock light would still react.
Sometimes, it doesn't, nor to num-lock, it won't even react to alt-sysrq-(resui)b...

I didn't find any kind of log info (journalctrl).

Before I forget to mention, I also tried to suspend with the mga5 kernel that was not replaced when upgrading to Mageia 6 - I installed the mga6 desktop kernel by hand. Suspend was just as broken.

My system: AMD A10 APU, 32GB Ram, SSD for the system, big magnetic drive for data / user accounts and lots of stuff via NFS from a server; 4k monitor on DP/display port and 720p projector (which is switched off, connected and visible to the system on VGA).

Any thoughts, hints?
Thanks in advance!

[doktor5000]

At least for pm-suspend you can enable debug logging. As root run

Code: Select all

PM_DEBUG=true pm-suspend

after resuming again (or after a reboot, as your system does not seem to allow any login) then the log of that should end up in /var/log/pm-suspend.log
For more information about this, see e.g. https://wiki.ubuntu.com/DebuggingKernelSuspend or maybe https://01.org/blogs/rzhang/2015/best-p ... ate-issues and maybe https://www.kernel.org/doc/Documentatio ... ugging.txt

For suspend via systemd, you could only attach the full journal of the last boot, that is after you rebooted when resuming did not work. So as root

Code: Select all

journalctl -ab-1 > /tmp/journal.log

and attach /tmp/journal.log here.

I believe this is more of a hardware / driver issue, so we would need more details where it actually hangs.

[artee]

Hey Doc, thanks for your reply!

I guess I had already found some of that at the same time, didn't have time to go into it so far, but your answer motivated me to give things a quick look.

I ran the commands for the pm_suspend debugging, the log file is attached. Nothing that I could see in there that gave me a strong hint...
After waking up the system, it came to the point where capslock doesn't, numlock does, and alt-sysrq-b also does. At least I didn't have to reach for the reset button on the pc...

I went through journalctl logs earlier today as well, nothing after the suspend/sleep comment, so the system is quite dead at that point.

I think I may be able to take some time tomorrow (time's up for tonight) to look into this further.

For now, my thoughts:
driver or hardware issue: this pc has worked with Mageia 4 and 5 without any problem with suspend to RAM, and I also tried with the Mageia 5 kernel that the upgrade of Mageia 6 kindly left on the system and in the bootloader.
Perhaps a live-dvd image will make it easier to debug...? Will download one anyway, if only to torrent out.
Or maybe I should have a look at all the rpmnew files that are scattered around the system since the upgrade...?

/me out
Gutenacht!

[artee]

Couldn't help myself, just had to try stuff with the live cd.

Result: suspend to RAM works fine.

Strangely enough, the live cd also doesn't shut down properly, gets a kernel panic, like the currently installed system...

[doktor5000]

Hmmm, could it be that the SSD itself is not waking up again? I mean the kernel itself still responds to a point where you can reboot, so the machine is halfway alive.

What mainboard do you use, and how is the SSD connected? NVMe? M.2? SATA? ... ?

[artee]

Hi doctor5000,
thanks again for thinking with me here...

I figured, perhaps I can find some bliss if I just install from DVD...

Then I run into a bug that looks like this one: https://bugs.mageia.org/show_bug.cgi?id=19935

you must have a bios partition for non-uefi gpt-partitioned disks

I have a spare partition to function as alternative root, so no repartitioning required.
This system has only gotten single use as Mageia, starting at 5, now upgraded to 6.
I do not want to delete/format the 300MB FAT partition, that way I may kill this system...

Related to the hardware:
One SSD (Samsung EVO 850? SATA in any case) of 250 GB, one 4TB rotating drive, SATA as well. Asus a88x Mainboard
Both drives are with GPT table (as recommended for Mageia 5), the SSD has /, /home, /mnt/altroot as EXT4, one swap and one /boot/EFI vfat partition.

So upon reading more I recalled that there was something about booting a disc and UEFI, and after setting the UEFI of the pc to boot the DVD UEFI style, I get into the UEFI installer.
That got rid of warning about the bios partition for non-uefi gpt-partitioned drives.

Instead, I got:

You must have a ESP FAT32 partition mounted in /boot/EFI

I recalled (after quite some messing about) with all the crashes, that at some point I did see a warning about the sda2 (/boot/EFI) partition needing file system checks.
So I did that (and stupidly confirmed that the hanging suspend does cause that..) using fsck, and booted to the dvd again, UEFI style.
Unfortunately, I still get:

You must have a ESP FAT32 partition mounted in /boot/EFI

I can mount it in the installer, on the console (alt-f3?) I can ls the one file in it, but if I choose view, the installer throws an error (not able to mount or so..?)...

Time for bed again...

[doktor5000]

Well, at least regarding to the motherboard or the AMD chipset on top there was not much I found, apart from issues with the autosuspend of the USB3 controller, see https://bugzilla.redhat.com/show_bug.cgi?id=1196943
Although what you definitely should check is if there's a BIOS update for yours. But I don't see A88X listed on the Asus page, only those: A88X-PRO, A88X-PLUS/USB 3.1, A88XM-PLUS, A88XM-A, A88XM-A/USB 3.1, A88X-PLUS, A88XM-E, A88XM-E/USB 3.1

Regarding the installation, from checking the bug report, the warning about the bios partition for non-uefi gpt-partitioned drives is bogus and can be ignored from what I can tell.

[artee]

Thanks again doktor5000!

As for the USB, I may havve come across that info, not sure if I had to do something (or if I did something), will have to check.

For now, the system works in a usable way, so until the weekend, or even after, I won't have time to try out stuff.

[artee]

I guess I'm not a man of my word today...

I found some info on the vfat partition needing to have the ESP and boot flags set.

Mine had the flag "msftdata" set, as soon as I opted for ESP with gparted, it also selected "boot"... so I just had to try, lo and behold, the installer is now runnung fine.

Strange that Mageia 5 never had problems to boot... it was the first OS installed on this box.


Edit:
I booted into the fresh install, Suspend to RAM is not working.
Same as on the upgraded system....
Now to figure out what is different for a live dvd vs an installed system.

[artee]

Well, I put the topic to solved, but that's only partially true...

Updated the BIOS/UEFI from v 1602 (2014) to v 2603 (2016, April) on my Asus A88X-Pro motherboard.
Found some things in the BIOS that are sure not to be as required, so configured those.

Next thing, the bootsplash doesn't show. Back into the BIOS. I guess all settings had reverted to failsafe or some such.

Configured the BIOS as I think it should be (fan speeds slow, nothing aggressive, all timings to safe or default or auto - overclocking has rarely proven useful to me). Reboot.

At graphical login - Ctrl-Alt-F2 to switch to CLI-login. As root, pm_suspend. Waking up, the prompt comes back - which it never did before. Good! Ctrl-Alt-F1, graphical login still alive. Logging in, plasma in view, Ctrl-Alt-F2 to switch to CLI, again pm_suspend. Works again. Braver, back to plasma, user suspend via menu.
Wake up - system works again! Hooray!

The (somewhat) bad news: I don't really know what did this...
Setting the S4 state to "enabled" or one of the many other options I tweaked in the UEFI/BIOS?
Updating the BIOS?

The somewhat worse bad news: Suspend-to-Disk now fails to work...
I took it out of the menu (by editing /etc/polkit-1/rules.d/10-disable-suspend.rules as per instructions somewhere on the WWW).

The really sad part: we'll never know why Suspend-to-RAM worked fine with Mageia 5 and the Mageia 6 live DVD, but didn't with an upgraded system nor with a clean installation.

The really good part: I will leave this behind me, since I got the mode that I wanted. Which is also why I put the topic to "solved".

Thanks again for thinking with me, doktor5000!
Herzlichen Dank!

[doktor5000]

The (somewhat) bad news: I don't really know what did this...
Setting the S4 state to "enabled" or one of the many other options I tweaked in the UEFI/BIOS?
Updating the BIOS?

Well, only you know what you tweaked or configured in particular ...

[artee]

Well, only you know what you tweaked or configured in particular ...

That's just it, I think I put everything back the way it was, but evidently, I haven't. In other words, no, I don't know.
Ah well... life goes on.

[doktor5000]

Yep it does

Please mark the thread accordingly by editing the topic of the first post and prefix it by [SOLVED], thanks