VioScsi Locking bug. 15e64ace .

[MartinCHarvey-Nutanix]

If no-one is working on the locking bug in vioscsi introduced by 15e64ac, I'll take a look. Not promising a rapid resolution: it's one of many things on my plate, and I'm not in the office for a chunk of this/next month.

MH.

[vrozenfe]

@MartinCHarvey-Nutanix
You are welcome to look into this issue.
However, in my opinion, it might be better to revert to the latest
stable code line and then review each new PR individually.

This situation with the code instability is obviously my fault for allowing it to happen.

There are some upcoming changes related to IOMME, DMA v3, and multiqueue support
that should be considered. I would prefer to build and test it on top of a baseline code that
has proven to be stable.

Best,
Vadim.

[benyamin-codez]

This situation with the code instability is obviously my fault for allowing it to happen.

It takes two to tango... 8^d
I had made some - perhaps unreasonable - assumptions about the QE workflow that I thought would have caught these.

I've got no problem owning up to, and fixing my mistakes - if I know about them...
What are the other known instabilities that need to be addressed...?
I'm presently reviewing my work here, so insights would be helpful.

I think viostor is sorted, yes?
In vioscsi, there's a nasty inline function tracing bug. I'm working on a quick PR for that.
Anything else?

@JonKohler, thanks for earlier bringing this one to my attention.

[MartinCHarvey-Nutanix]

Vadim, Thanks for your prompt reply. I'll put a patch out when we've finished some in-house verification testing, and I've got back from PTO. MH.

[MartinCHarvey-Nutanix]

Now submitted PR in lieu of patch.

[benyamin-codez]

PR #1293 and thank you Martin for your efforts on this.