[SOLVED] Strange Display-centric Problem

[Jagadguru]

Hi Guys,

 

System info in my signature. I'm having the strange problem that performance in my Windows VM goes to utter stuttering when and only when I turn my monitor off (and then back on). Performance is great until I do that. This VM is only for productivity, no gaming. Restarting the VM does not help. The only way to bring performance back to normal it to RESTART THE WHOLE "SERVER"

 

My MacOS VM works great always on the same GPU's passed through.

 

Is this some kind of graphics reset problem?

 

Have tried with nvidia driver and windows driver. Have tried with Q35 and i440fx machine types.

 

Also, There are dmesg errors:

 

Apr 4 17:53:08 Tower kernel: irq 16: nobody cared (try booting with the "irqpoll" option)
Apr 4 17:53:08 Tower kernel: CPU: 2 PID: 0 Comm: swapper/2 Not tainted 4.19.107-Unraid #1
Apr 4 17:53:08 Tower kernel: Hardware name: Dell Inc. OptiPlex 9020/06X1TJ, BIOS A18 03/13/2017
Apr 4 17:53:08 Tower kernel: Call Trace:
Apr 4 17:53:08 Tower kernel: <IRQ>
Apr 4 17:53:08 Tower kernel: dump_stack+0x67/0x83
Apr 4 17:53:08 Tower kernel: __report_bad_irq+0x30/0xa5
Apr 4 17:53:08 Tower kernel: note_interrupt+0x1d8/0x229
Apr 4 17:53:08 Tower kernel: handle_irq_event_percpu+0x4f/0x6f
Apr 4 17:53:08 Tower kernel: handle_irq_event+0x34/0x51
Apr 4 17:53:08 Tower kernel: handle_fasteoi_irq+0x92/0xfc
Apr 4 17:53:08 Tower kernel: handle_irq+0x1c/0x1f
Apr 4 17:53:08 Tower kernel: do_IRQ+0x46/0xd0
Apr 4 17:53:08 Tower kernel: common_interrupt+0xf/0xf
Apr 4 17:53:08 Tower kernel: </IRQ>
Apr 4 17:53:08 Tower kernel: RIP: 0010:cpuidle_enter_state+0xe8/0x141
Apr 4 17:53:08 Tower kernel: Code: ff 45 84 f6 74 1d 9c 58 0f 1f 44 00 00 0f ba e0 09 73 09 0f 0b fa 66 0f 1f 44 00 00 31 ff e8 7a 8d bb ff fb 66 0f 1f 44 00 00 <48> 2b 2c 24 b8 ff ff ff 7f 48 b9 ff ff ff ff f3 01 00 00 48 39 cd
Apr 4 17:53:08 Tower kernel: RSP: 0018:ffffc900031b3e98 EFLAGS: 00000246 ORIG_RAX: ffffffffffffffdc
Apr 4 17:53:08 Tower kernel: RAX: ffff88881e91fac0 RBX: ffff88881e92a300 RCX: 000000000000001f
Apr 4 17:53:08 Tower kernel: RDX: 0000000000000000 RSI: 0000000026e0b25e RDI: 0000000000000000
Apr 4 17:53:08 Tower kernel: RBP: 0000a02a208fd374 R08: 0000a02a208fd374 R09: 0000000000000001
Apr 4 17:53:08 Tower kernel: R10: 0000000000000000 R11: 071c71c71c71c71c R12: 0000000000000001
Apr 4 17:53:08 Tower kernel: R13: ffffffff81e5b120 R14: 0000000000000000 R15: ffffffff81e5b198
Apr 4 17:53:08 Tower kernel: ? cpuidle_enter_state+0xbf/0x141
Apr 4 17:53:08 Tower kernel: do_idle+0x17e/0x1fc
Apr 4 17:53:08 Tower kernel: cpu_startup_entry+0x6a/0x6c
Apr 4 17:53:08 Tower kernel: start_secondary+0x197/0x1b2
Apr 4 17:53:08 Tower kernel: secondary_startup_64+0xa4/0xb0
Apr 4 17:53:08 Tower kernel: handlers:
Apr 4 17:53:08 Tower kernel: [<00000000a8447878>] usb_hcd_irq
Apr 4 17:53:08 Tower kernel: [<00000000b4f21068>] vfio_intx_handler

Edited April 4, 2020 by Jagadguru

[Jagadguru]

Added dmesg error above.

[Jagadguru]

SOLVED by putting irqpoll=1 in the linux start line.

[Jagadguru]

UPDATE: Turned on MSI interrupts in Windows and removed irqpoll=1 from unRAID boot line. Still get "IRQ16 Nobody cared...IRQ Disabled" When I plug in the monitor by DisplayPort But performance does not tank.