https://bugzilla.kernel.org/show_bug.cgi?id=201631 --- Comment #33 from Benjamin Herrenschmidt (benh@xxxxxxxxxxxxxxxxxxx) --- Jan, For KVM, it can... you can back the guest using a file, and there's some interest in doing just that to back guests using large "new memory" pmem stuff for example. This actually leads to more generic variant of that issue which is the need to be able to provide proper "pinned" memory for some adapters in guest even if they use ATSD, as not all adapters can deal with faults on everything (for example they might be able to deal with fault on user buffers but not the descriptor ring). A discussion for another day I suppose.. As for AMD, I was looking at attachment https://bugzilla.kernel.org/attachment.cgi?id=280057 posted in comment 18 -- You are receiving this mail because: You are watching the assignee of the bug.
