VirtualBox 5.0 causes lost interrupts / disk stalls on a Linux system

[znmebznmeb]

The host system is an 8-core AMD 64-bit system with 32 GB of RAM, a 128 GB SSD drive and a 1 TB regular rotating disk drive. The operating system is Linux Mint 18; kernel is Ubuntu's

Linux AlgoCompSynth 4.4.0-45-generic #66-Ubuntu SMP Wed Oct 19 14:12:37 UTC 2016 x86_64 x86_64 x86_64 GNU/Linux

VirtualBox is 5.0.28 r111378 directly from Oracle's download page. I've tried this with 5.1 and it's worse.

In normal use - creating a virtual machine, exporting to an OVA, etc., sometimes the system will freeze for a second or two, then pick up where it left off. It only does this when I'm using VirtualBox.

I dug through kernel "dmesg" output and found out what's happening.

[  +0.418779] vboxdrv: ffffffffc0835020 VMMR0.r0
[  +0.108599] vboxdrv: ffffffffc0930020 VBoxDDR0.r0
[  +0.003241] vboxdrv: ffffffffc0006020 VBoxDD2R0.r0
[  +0.056406] vboxpci: created IOMMU domain ffff8808170fec20
[Nov 6 22:38] ata1: lost interrupt (Status 0x58)
[  +0.000029] ata1.01: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x6 frozen
[  +0.000007] ata1.01: failed command: READ DMA
[  +0.000009] ata1.01: cmd c8/00:01:00:e8:d9/00:00:00:00:00/f7 tag 0 dma 512 in
                       res 40/00:00:00:4f:c2/00:00:00:00:00/10 Emask 0x4 (timeout)
[  +0.000005] ata1.01: status: { DRDY }
[  +0.000023] ata1: soft resetting link
[  +0.184530] ata1.00: configured for UDMA/33
[  +0.007848] ata1.01: configured for UDMA/33
[  +0.000018] ata1: EH complete
[  +0.019241] ata1.00: Enabling discard_zeroes_data
[  +0.156301] ata1.00: Enabling discard_zeroes_data

I've seen this happen on both the SSD and the rotating disk - whichever VirtualBox is using at the time.

[znmebznmeb]

VBox.log from the session - I'm not sure about the timestamps since dmesg shows local time (PST) and your log looks like UTC