From d7def8317a598329ad6e520e57984ca5feac6ede Mon Sep 17 00:00:00 2001 From: Gabriela Cervantes Date: Tue, 26 Sep 2023 16:30:59 +0000 Subject: [PATCH] metrics: Fix general check static warnings This PR fixes general check static warnings. Signed-off-by: Gabriela Cervantes --- docs/design/kata-vra.md | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/docs/design/kata-vra.md b/docs/design/kata-vra.md index 887f51aec..bc62e762c 100644 --- a/docs/design/kata-vra.md +++ b/docs/design/kata-vra.md @@ -43,7 +43,7 @@ and perform DMA transactions _anywhere_. The second feature is ACS (Access Control Services), which controls which devices are allowed to communicate with one another and thus avoids improper -routing of packets irrespectively of whether IOMMU is enabled or not. +routing of packets `irrespectively` of whether IOMMU is enabled or not. When IOMMU is enabled, ACS is normally configured to force all PCI Express DMA to go through the root complex so IOMMU can translate it, impacting performance @@ -126,7 +126,7 @@ efficient P2P communication. ## PCI Express Virtual P2P Approval Capability Most of the time, the PCI Express topology is flattened and obfuscated to ensure -easy migration of the VM image between different physical hardware topologies. +easy migration of the VM image between different physical hardware `topologies`. In Kata, we can configure the hypervisor to use PCI Express root ports to hotplug the VFIO  devices one is passing through. A user can select how many PCI Express root ports to allocate depending on how many devices are passed through. @@ -300,7 +300,7 @@ pcie_switch_port = 8 ``` Each device that is passed through is attached to a PCI Express downstream port -as illustrated below. We can even replicate the host’s two DPUs topologies with +as illustrated below. We can even replicate the host’s two DPUs `topologies` with added metadata through the CDI. Most of the time, a container only needs one pair of GPU and NIC for GPUDirect RDMA. This is more of a showcase of what we can do with the power of Kata and CDI. One could even think of adding groups of @@ -328,7 +328,7 @@ $ lspci -tv ``` The configuration of using either the root port or switch port can be applied on -a per Container or Pod basis, meaning we can switch PCI Express topologies on +a per Container or Pod basis, meaning we can switch PCI Express `topologies` on each run of an application. ## Hypervisor Resource Limits