Per-entrypoint DevMaps and Mailboxes
#710
Draft
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
We still have a hilarious failure to remove resources *somewhere*, causing havoc on underlay removal.
FelixMcFelix
force-pushed
the
per-entry-devmap
branch
from
cafcfb5 to
527d89f
Compare
We can now batch all packets up to the intended caller (!). Looking to be around 16Gbps on `glasgow` zone-to-zone.
FelixMcFelix
changed the title
DevMapsDevMaps and Mailboxes
Well, that simplifies things!
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR provides every possible entrypoint with its own (mutex-wrapped) context, so that the Tx and Rx paths no longer contend over a single
KRwLock. The role of these contexts is to provide a port lookup map for delivery to local ports in both underlay rx and loopback. These entrypoints are one per OPTE port (in the tx case), and one for every CPU on every underlay link (as we don't know the full set of thread bindings who might call into us). Addition/removal of a port then goes through all entrypoints one by one and updates them in turn.The very useful part of this is that we have a convenient place to build and store chains of packets bound toward a single sender -- and on top of the
siphonbranch, we do have sets of packets needing batched up for delivery to mac. The Rx path now first deposits processed packets into chains by target port (keyed on MAC+VNI). Onglasgowthis looks to get us 16Gbps single stream (with no bump to aggregate/-P2and upwards throughput).Closes #689.