NVMe drive stops working after rmmod/modprobe nvme => fixed in SVN/5.1.x

[Sven]

How to reproduce

• boot ubuntu 16.10 live CD
• dd if=/dev/nvme0n1 of=/dev/null works as expected
• rmmod nvme
• modprobe nvme
• dd if=/dev/nvme0n1 of=/dev/null fails with an error

dmesg shows the following errors: nvme nvme0: Device not ready; aborting initialisation nvme nvme0: Removing after probe failure status: -19

[aeichner]

Thanks for the report, this will be fixed in the next maintenance release.

[Frank Mehnert]

The fix is part of the most recent 5.1 Extension Pack (>= 112114) which you can download here.

[Sven]

I just checked with 5.1.12 and rmmod nvme followed by modprobe nvme now works. I can read the disk after even after rmmod/modprobe.

However, rmmod nvme takes a long time to complete. More precisely, it takes about 60 seconds. It seems as if the driver is waiting for some response - as if the module could only be removed when some internal driver timeout occurs.

[Frank Mehnert]

Still reproducible with VBox 5.1.18? I couldn't. Using an Ubuntu 16.10 guest with 1 disk attached to the NVMe controller. The nvme module is automatically loaded (together with nvme_core). The disk is accessible by /dev/nvme0n1. I could also rmmod nvme + nvme nvme_core + modprobe nvme. No problem either. Please reopen if this is still a problem for you. In that case, please attach a VBox.log file of such a VM session (mainly to see if your VM configuration is somehow special) and provide exact steps to reproduce.