|Priority:||4 - Normal|
|Created by:||Patrick Mooney [X]|
|Reported by:||Patrick Mooney [X]|
|Assigned to:||Patrick Mooney [X]|
Fixed: A fix for this issue is checked into the tree and tested.
(Resolution Date: 2019-02-14T15:28:39.364Z)
2019-02-28 Mind Grapes (Release Date: 2019-02-28)
The bhyve VM shim which implements EPT should be implemented in terms of the generic page table API proposed in OS-7437. Ideally, as much logic as possible would be shared between EPT and NPT, once the AMD side is ported over.
Testing for this consisted of bringing various guests up and down on a hypervisor running the updated platform image. Under both DEBUG and non-DEBUG kernels, no issues were observed during EPT setup, operation, or tear-down. This was attempted with fully "wired" memory (via the
-S arg to bhyve) in addition to normal operation. Guests sized both smaller (512MB) and larger (8+GB) than the 1G highmem limit were tested. After a series of tests on the DEBUG kernel, it was stopped with
halt -d and the subsequent dump was checked for leaks. None related to bhyve were found. (There was some existing noise from ACPI)
Hans performed some PCI pass-thru tests with that PI and observed no issues:
looks like it's working fine, the ppt VM started and the nvidia samples still work
Waiting until @ryan.kitchen runs a set of the QA tests on a platform bearing the change before this is merged.
illumos-joyent commit c8b89b2e0ef5e796d5ea1ee45802f7ebc39545f4 (branch master, by Patrick Mooney)
OS-7438 refactor bhyve EPT to use generic page tables
OS-7437 want generic indexed page table system
Reviewed by: John Levon <email@example.com>
Reviewed by: Hans Rosenfeld <firstname.lastname@example.org>
Approved by: Hans Rosenfeld <email@example.com>