Matt LaFayette (Kurjanowicz)

nvme: allow PCI ID overrides

8

I have found it useful to ask the nvme device to pretend to be a different PCI device type. This is useful for testing and prototyping. Validated via unit tests.

Reminding myself that I said I would do this. We have several maintainer functions (creating new branches, marking tasks, and other process stuff). Document this.

openhcl: servicing: tests from v.prev <-> v.current

4

From TODO in #445

@jstarks mentioned this in an offline conversation. He noticed that if you: 1. Establish a connection, 2. Drive IO 3. Wait several minutes (without IO) 4. Issue new IO The...

openhcl_boot: disable sidecar for smallish VMs on the restore path (#2438)

2

Clean cherry-pick of PR #2438. Empirical data shows that sidecar makes the particular situation addressed in this flow worse. So, disable sidecar if: - This is a servicing restore, -...

DO NOT MERGE: nvme_driver: delay io queue create to be post restore iff there are no ouststanding IO to that queue (#2456)

1

Clean cherry pick of PR #2456. Looking at production data, I found that restoring the IO queues took significantly longer than I expected. In those same runs, there was no...

RFC: sidecar: allow openhcl_boot to specify a state per-cpu, and have openhcl_boot set that state for VPs that have mapped interrupts

4

**DO NOT MERGE, TESTING IN PROGRESS** Experiment: ask the kernel to online any CPUs that will immediately get interrupts mapped, and ask sidecar to online the rest.

RFC: allow command line to specify controller + lun/nsid

1

I was surprised that adding additional uh-nvme disks didn't add multiple controllers, so add that capability. I lost steam when trying to hook this up to the openvmm storage builder,...