diff options
author | Oliver Upton <oliver.upton@linux.dev> | 2024-03-07 03:56:01 +0300 |
---|---|---|
committer | Oliver Upton <oliver.upton@linux.dev> | 2024-03-07 03:56:01 +0300 |
commit | 9bd8d7df1971c2ebdcaf4526cf7a3f4ea38d0ede (patch) | |
tree | 1a773f15a68befda40169b1a8432742909d2b3f3 /lib/test-string_helpers.c | |
parent | 8dbc41105e96641e9c1569f512d19f0046a02463 (diff) | |
parent | a39d3a966a090989b89c0287a67cd98c85ae2f52 (diff) | |
download | linux-9bd8d7df1971c2ebdcaf4526cf7a3f4ea38d0ede.tar.xz |
Merge branch kvm-arm64/vfio-normal-nc into kvmarm/next
* kvm-arm64/vfio-normal-nc:
: Normal-NC support for vfio-pci @ stage-2, courtesy of Ankit Agrawal
:
: KVM's policy to date has been that any and all MMIO mapping at stage-2
: is treated as Device-nGnRE. This is primarily done due to concerns of
: the guest triggering uncontainable failures in the system if they manage
: to tickle the device / memory system the wrong way, though this is
: unnecessarily restrictive for devices that can be reasoned as 'safe'.
:
: Unsurprisingly, the Device-* mapping can really hurt the performance of
: assigned devices that can handle Gathering, and can be an outright
: correctness issue if the guest driver does unaligned accesses.
:
: Rather than opening the floodgates to the full ecosystem of devices that
: can be exposed to VMs, take the conservative approach and allow PCI
: devices to be mapped as Normal-NC since it has been determined to be
: 'safe'.
vfio: Convey kvm that the vfio-pci device is wc safe
KVM: arm64: Set io memory s2 pte as normalnc for vfio pci device
mm: Introduce new flag to indicate wc safe
KVM: arm64: Introduce new flag for non-cacheable IO memory
Signed-off-by: Oliver Upton <oliver.upton@linux.dev>
Diffstat (limited to 'lib/test-string_helpers.c')
0 files changed, 0 insertions, 0 deletions