On Mon, Sep 14, 2026 at 03:06:18PM +0300, Leon Romanovsky wrote:
On Mon, Sep 14, 2026 at 08:58:26AM -0300, Jason Gunthorpe wrote:
On Mon, Sep 14, 2026 at 02:54:56PM +0300, Leon Romanovsky wrote:
So I still think that "temporarily revoked" is closer to invalidate_mappings, with some additional dma-buf documentation to lose the expectation that the buffer will become available again.
It really isn't, there is no expectation her that the importer has to put things back like with move.
As we discussed with Christian, "move" doesn't have such expectation. https://lore.kernel.org/all/afb19db0-4154-443c-9d39-f905e7d17806@amd.com/
"the exporter can move the buffer back when a mapping is created" is exactly what I'm saying is not required here, iommufd will not implement this, it only does permanent revoke.
Jason