On 3/7/20 10:29 PM, Christoph Hellwig wrote:

So I looked into this, and if it was just the uevent this
should have been fixed in:

"block: don't send uevent for empty disk when not invalidating"

from Eric Biggers in December.  Does the problem still occur with that
patch applied?

Yes, it occurs with that patch. The patch that introduces the issue and the one
you mentioned above both went in in v5.5-rc1. I found this issue in v5.5
release.

Thanks,
Zhe