Guest linux fails to receive key input after X11 screen appears on resume from saved state.

[ci-zephyurus]

Hi,

I experienced an apparent hung of guest linux inside virtualbox 5.2.20 r125813 running under Windows 10. After a saved state is resumed, the X11 screen is displayed and there are teminal windows shown on it. But when I try to type something to a consle window nothing gets echoed back. Come to think of it, USB mouse does not respond at all. I have to kill it (power down the guest OS).

Gesut linux version is Debian GNU/Linux kernel version: uname -a Linux ip030 4.16.0-2-amd64 #1 SMP Debian 4.16.16-2 (2018-06-22) x86_64 GNU/Linux

I have this apparent hung problem three times since I move the image from a XEON-based PC to AMD Ryzen 1700 PC about 10 days ago.

Is it possible that AMD Ryzen and guest linux may have some timing-related issue (like say, timer or counter not handled properly using hardware counter?)

I am asking because I also have a very starnage network adaptor hung (e1000 network card driver got hung). I have not seen this bug often during 5 or 6 years of using the image on a XEON-based PC.

While looking at dmesg output, I noticed that kernel says TSC is unstable or whatever: I am not sure what the message under XEON PC was.

 dmesg | grep -5 -i bogo 
[    0.004000] hpet clockevent registered
[    0.004000] APIC: Switch to symmetric I/O mode setup
[    0.004000] ..TIMER: vector=0x30 apic1=0 pin1=2 apic2=-1 pin2=-1
[    0.004005] tsc: Detected 2994.376 MHz processor
[    0.004007] tsc: Marking TSC unstable due to TSCs unsynchronized
[    0.004008] Calibrating delay loop (skipped) preset value.. 5988.75 BogoMIPS (lpj=11977504)
[    0.004010] pid_max: default: 32768 minimum: 301
[    0.004046] Security Framework initialized
[    0.004047] Yama: disabled by default; enable with sysctl kernel.yama.*
[    0.004060] AppArmor: AppArmor initialized
[    0.016000] Dentry cache hash table entries: 2097152 (order: 12, 16777216 bytes)
--
[    0.024084]  #6
[    0.004000] kvm-clock: cpu 6, msr 3:23fea181, secondary cpu clock
[    0.004000] mce: CPU supports 0 MCE banks
[    0.024443] smp: Brought up 1 node, 7 CPUs
[    0.024443] smpboot: Max logical packages: 1
[    0.024443] smpboot: Total of 7 processors activated (41921.26 BogoMIPS)
[    0.025189] devtmpfs: initialized
[    0.025189] x86/mm: Memory block size: 128MB
[    0.025209] clocksource: jiffies: mask: 0xffffffff max_cycles: 0xffffffff, max_idle_ns: 7645041785100000 ns
[    0.025209] futex hash table entries: 2048 (order: 5, 131072 bytes)
[    0.025209] pinctrl core: initialized pinctrl subsystem

Anyway, I am attaching Vbox.log when the guest does not receive any input and I had to kill (power down the image). Maybe there might be a clue in there.

[ci-zephyurus]

VBox.log when I had to kill the guest (power down the guest) after an obvious hung [no keyboard input accepted]

[ci-zephyurus]

The hung happens again. The gust linux is not responsive after resume.

In the log, I see

00:00:22.119767 GUI: UIMachineLogicNormal::sltCheckForRequestedVisualStateType: Requested-state=0, Machine-state=5
00:00:25.629737 VMMDev: vmmDevHeartbeatFlatlinedTimer: Guest seems to be unresponsive. Last heartbeat received 4 seconds ago

So the guest linux OS is stuck...

I am attaching the log from this hung state again. I wonder if there is any tell-tale sign in it.

I am beginning to wonder if there may be Ryzen-specific issue (?).

[ci-zephyurus]

Log: resuming saved linux64 guest when the guest fails to be responsive.