nvme: Add optimizations for device removal
AbandonedPublic
Actions

Authored by imp on Feb 11 2021, 1:05 AM.

Details

Reviewers

mav
chs
chuck

Summary

When a device is hot unplugged (surprise removal) we can wind up in the recovery
path. When we detect the device is gone in the recovery path, short-circuit the
rest of recovery and fail the controller.

Diff Detail

Repository

rS FreeBSD src repository - subversion

Lint

Lint Passed

Unit

No Test Coverage

Build Status

Buildable 36882
Build 33771: arc lint + arc unit

Event Timeline

imp created this revision.Feb 11 2021, 1:05 AM

Herald added a subscriber: dab. · View Herald TranscriptFeb 11 2021, 1:05 AM

imp requested review of this revision.Feb 11 2021, 1:05 AM

Harbormaster completed remote builds in B36882: Diff 83671.Feb 11 2021, 1:05 AM

imp added reviewers: mav, chs, chuck.Feb 11 2021, 1:05 AM

From past experience, I've noticed that HGST/WDC drives set CFS on power loss before being NVME_GONE. In this case, would the state machine recover as it doesn't check NVME_GONE again?

In D28585#640561, @chuck wrote:

From past experience, I've noticed that HGST/WDC drives set CFS on power loss before being NVME_GONE. In this case, would the state machine recover as it doesn't check NVME_GONE again?

We check in other places in the driver and should recover.

I've not, however, tested that since I have no real means to do so short of turning off a PHY on a carrier card we us which isn't quite the same thing.

OBE. Even when I had this in, it only helped a little and failure is more nuanced than this patch contemplates.