On Fri, Nov 12, 2021 at 9:08 AM Sagi Grimberg [off-list ref] wrote:

quoted

quoted

Completely disagree here. The TCP original report was just an example of
lack of protection we have against spurious completions. Nothing
specific about nvme-tcp here, this was discussed and agreed on in
the original report.

You are ignoring the facts:

1. The device that broke the spec in the first place was that device for
which caused you to add the gen bits to CID.

Correct.

Can we also acknowledge that the broken device isn't really nvme? Apple uses
PCI class code 018020, which is an unspecified mass storage controller, clearly
*not* nvme. Apple does not care if their devices work with a generic
NVMe driver.
The only reason it works in Linux is because (1) a savvy user base exists to
reverse engineer their crap, and (2) their crap so far hasn't been invasive
enough to justify an entirely different driver. No one should consider Apple an
example to follow.