Integrate SoftNPU as virtual hardware #2089

internet-diglett · 2022-12-21T22:24:44Z

Closes #1465
Closes oxidecomputer/opte#236

tools/scrimlet/softnpu-init.sh

docs/boundary-services-a-to-z.adoc

rcgoodfellow · 2023-02-11T00:24:56Z

Just dropping a note that I was able to successfully test this end-to-end with the CI builds brought in via the setup-virtual-hardware scripts. From zero to instances talking to the Internet.

rcgoodfellow · 2023-02-11T00:30:53Z

Looks like destroy virtual hardware needs a bit of love.

ry@masaka:~/src/omicron$ pfexec ./tools/destroy_virtual_hardware.sh
+++ dirname ./tools/destroy_virtual_hardware.sh
++ cd ./tools
++ pwd
+ SOURCE_DIR=/home/ry/src/omicron/tools
+ cd /home/ry/src/omicron/tools/..
+ OMICRON_TOP=/home/ry/src/omicron
+ MARKER=/etc/opt/oxide/NO_INSTALL
+ [[ -f /etc/opt/oxide/NO_INSTALL ]]
++ id -u
+ [[ 0 -ne 0 ]]
+ verify_omicron_uninstalled
+ svcs svc:/system/illumos/sled-agent:default
svcs: Pattern 'svc:/system/illumos/sled-agent:default' doesn't match any instances
+ unload_xde_driver
++ modinfo
++ grep xde
++ cut -d ' ' -f 1
+ local ID=239
+ [[ -n 239 ]]
+ modunload -i 239
+ success 'Verified the xde kernel driver is unloaded'
+ set +x
Verified the xde kernel driver is unloaded
+ try_remove_vnics
+ try_remove_address lo0/underlay
+ local ADDRESS=lo0/underlay
++ ipadm show-addr -p -o addr lo0/underlay
ipadm: Address object not found+ [[ -n '' ]]
+ success 'Verified address lo0/underlay does not exist'
+ set +x
Verified address lo0/underlay does not exist
+ try_remove_interface sc0_1
+ local IFACE=sc0_1
++ ipadm show-if -p -o IFNAME sc0_1
ipadm: Could not get interface(s): Interface does not exist+ [[ -n '' ]]
+ success 'Verified IP interface sc0_1 does not exist'
+ set +x
Verified IP interface sc0_1 does not exist
+ try_remove_vnic sc0_1
+ local LINK=sc0_1
++ dladm show-vnic -p -o LINK sc0_1
+ [[ -n sc0_1 ]]
+ dladm delete-vnic sc0_1
dladm: vnic deletion failed: link busy
+ warn 'Failed to delete VNIC link sc0_1'
+ set +x
Failed to delete VNIC link sc0_1
+ success 'Verified VNIC link sc0_1 does not exist'
+ set +x
Verified VNIC link sc0_1 does not exist
+ INDICES=("0" "1")
+ for I in "${INDICES[@]}"
+ try_remove_simnet net0
+ local LINK=net0
++ dladm show-simnet -p -o LINK net0
+ [[ -n net0 ]]
+ dladm delete-simnet -t net0
dladm: simnet deletion failed: link busy
+ warn 'Failed to delete simnet link net0'
+ set +x
Failed to delete simnet link net0
+ success 'Verified simnet link net0 does not exist'
+ set +x
Verified simnet link net0 does not exist
+ try_remove_simnet sc0_0
+ local LINK=sc0_0
++ dladm show-simnet -p -o LINK sc0_0
+ [[ -n sc0_0 ]]
+ dladm delete-simnet -t sc0_0
dladm: simnet deletion failed: link busy
+ warn 'Failed to delete simnet link sc0_0'
+ set +x
Failed to delete simnet link sc0_0
+ success 'Verified simnet link sc0_0 does not exist'
+ set +x
Verified simnet link sc0_0 does not exist
+ try_remove_simnet sr0_0
+ local LINK=sr0_0
++ dladm show-simnet -p -o LINK sr0_0
+ [[ -n sr0_0 ]]
+ dladm delete-simnet -t sr0_0
dladm: simnet deletion failed: link busy
+ warn 'Failed to delete simnet link sr0_0'
+ set +x
Failed to delete simnet link sr0_0
+ success 'Verified simnet link sr0_0 does not exist'
+ set +x
Verified simnet link sr0_0 does not exist
+ try_remove_simnet scr0_0
+ local LINK=scr0_0
++ dladm show-simnet -p -o LINK scr0_0
+ [[ -n scr0_0 ]]
+ dladm delete-simnet -t scr0_0
+ success 'Verified simnet link scr0_0 does not exist'
+ set +x
Verified simnet link scr0_0 does not exist
+ for I in "${INDICES[@]}"
+ try_remove_simnet net1
+ local LINK=net1
++ dladm show-simnet -p -o LINK net1
+ [[ -n net1 ]]
+ dladm delete-simnet -t net1
dladm: simnet deletion failed: link busy
+ warn 'Failed to delete simnet link net1'
+ set +x
Failed to delete simnet link net1
+ success 'Verified simnet link net1 does not exist'
+ set +x
Verified simnet link net1 does not exist
+ try_remove_simnet sc1_0
+ local LINK=sc1_0
++ dladm show-simnet -p -o LINK sc1_0
+ [[ -n sc1_0 ]]
+ dladm delete-simnet -t sc1_0
+ success 'Verified simnet link sc1_0 does not exist'
+ set +x
Verified simnet link sc1_0 does not exist
+ try_remove_simnet sr0_1
+ local LINK=sr0_1
++ dladm show-simnet -p -o LINK sr0_1
+ [[ -n sr0_1 ]]
+ dladm delete-simnet -t sr0_1
dladm: simnet deletion failed: link busy
+ warn 'Failed to delete simnet link sr0_1'
+ set +x
Failed to delete simnet link sr0_1
+ success 'Verified simnet link sr0_1 does not exist'
+ set +x
Verified simnet link sr0_1 does not exist
+ try_remove_simnet scr0_1
+ local LINK=scr0_1
++ dladm show-simnet -p -o LINK scr0_1
+ [[ -n scr0_1 ]]
+ dladm delete-simnet -t scr0_1
+ success 'Verified simnet link scr0_1 does not exist'
+ set +x
Verified simnet link scr0_1 does not exist
+ try_destroy_zpools
+ readarray -t ZPOOLS
++ zfs list -d 0 -o name
++ grep '^oxp_'
+ for ZPOOL in "${ZPOOLS[@]}"
+ VDEV_FILE=/home/ry/src/omicron/oxp_d462a7f7-b628-40fe-80ff-4e4189e2d62b.vdev
+ zfs destroy -r oxp_d462a7f7-b628-40fe-80ff-4e4189e2d62b
+ zfs unmount oxp_d462a7f7-b628-40fe-80ff-4e4189e2d62b
+ zpool destroy oxp_d462a7f7-b628-40fe-80ff-4e4189e2d62b
+ rm -f /home/ry/src/omicron/oxp_d462a7f7-b628-40fe-80ff-4e4189e2d62b.vdev
+ success 'Verified ZFS pool and vdev oxp_d462a7f7-b628-40fe-80ff-4e4189e2d62b does not exist'
+ set +x
Verified ZFS pool and vdev oxp_d462a7f7-b628-40fe-80ff-4e4189e2d62b does not exist
+ for ZPOOL in "${ZPOOLS[@]}"
+ VDEV_FILE=/home/ry/src/omicron/oxp_e4b4dc87-ab46-49fb-a4b4-d361ae214c03.vdev
+ zfs destroy -r oxp_e4b4dc87-ab46-49fb-a4b4-d361ae214c03
+ zfs unmount oxp_e4b4dc87-ab46-49fb-a4b4-d361ae214c03
+ zpool destroy oxp_e4b4dc87-ab46-49fb-a4b4-d361ae214c03
+ rm -f /home/ry/src/omicron/oxp_e4b4dc87-ab46-49fb-a4b4-d361ae214c03.vdev
+ success 'Verified ZFS pool and vdev oxp_e4b4dc87-ab46-49fb-a4b4-d361ae214c03 does not exist'
+ set +x
Verified ZFS pool and vdev oxp_e4b4dc87-ab46-49fb-a4b4-d361ae214c03 does not exist
+ for ZPOOL in "${ZPOOLS[@]}"
+ VDEV_FILE=/home/ry/src/omicron/oxp_f4b4dc87-ab46-49fb-a4b4-d361ae214c03.vdev
+ zfs destroy -r oxp_f4b4dc87-ab46-49fb-a4b4-d361ae214c03
+ zfs unmount oxp_f4b4dc87-ab46-49fb-a4b4-d361ae214c03
+ zpool destroy oxp_f4b4dc87-ab46-49fb-a4b4-d361ae214c03
+ rm -f /home/ry/src/omicron/oxp_f4b4dc87-ab46-49fb-a4b4-d361ae214c03.vdev
+ success 'Verified ZFS pool and vdev oxp_f4b4dc87-ab46-49fb-a4b4-d361ae214c03 does not exist'
+ set +x
Verified ZFS pool and vdev oxp_f4b4dc87-ab46-49fb-a4b4-d361ae214c03 does not exist
+ remove_softnpu_zone
+ zoneadm -z softnpu halt
+ zoneadm -z softnpu uninstall -F
INFO: omicron: uninstalling zone softnpu...
+ zonecfg -z softnpu delete -F
+ rm -rf /opt/softnpu/stuff
+ zfs destroy rpool/softnpu-zone
+ rm -rf /softnpu-zone
ry@masaka:~/src/omicron$ ipadm
ADDROBJ           TYPE     STATE        ADDR
lo0/v4            static   ok           127.0.0.1/8
igb0/dhcp         dhcp     ok           192.168.1.4/24
lo0/v6            static   ok           ::1/128
igb0/v6           addrconf ok           fe80::aa5e:45ff:fecd:bd94/10
igb0/v6           addrconf ok           2603:3024:b0b:3400::18ba/128
ry@masaka:~/src/omicron$ ipadm show-if
IFNAME     CLASS     STATE    CURRENT      PERSISTENT
lo0        VIRTUAL   ok       -m-v------46 ---
igb0       IP        ok       bm--------46 -46
net0       IP        down     bm--------46 ---
net1       IP        down     bm--------46 ---
ry@masaka:~/src/omicron$ dladm
LINK        CLASS     MTU    STATE    BRIDGE     OVER
igb0        phys      1500   up       --         --
net0        simnet    1600   up       --         sc0_0
sc0_0       simnet    1600   up       --         net0
sr0_0       simnet    9000   up       --         --
net1        simnet    1600   up       --         --
sr0_1       simnet    9000   up       --         --
sc0_1       vnic      1500   up       --         igb0

rcgoodfellow · 2023-02-24T04:30:07Z

Looks like destroy virtual hardware needs a bit of love.

This is now working as expected for me.

rcgoodfellow · 2023-02-24T05:05:12Z

I sorted out the networking issues we were seeing today with softnpu in the following

And pushed an Omicron commit you can cherry-pick into this branch with the updates.

428a035

rcgoodfellow · 2023-02-26T21:56:14Z

Noting that we'll need to update to the current Dendrite API as we are now a few commits behind. This is causing saga failure for me due to API mismatch.

https://github.com/oxidecomputer/dendrite/pull/87

rcgoodfellow · 2023-02-27T06:09:45Z

I've put together a pile of pull requests to fix softnpu/p4 range keys.

P4 more shoring up of endianness p4#20
Softnpu endian updates softnpu#5
Sidecar-lite endian updates sidecar-lite#8
Dendrite https://github.com/oxidecomputer/dendrite/pull/104

These updates collectively solve the issue I was seeing with proxy arp replying to a much larger range than assigned. I think they may also address the NAT range issues @internet-diglett has been observing.

bnaecker

Not at all a complete review. Just a few notes that came up during chat today, wanted to get them persisted.

bnaecker · 2023-03-02T22:16:40Z

nexus/src/app/sagas/instance_create.rs

+    let params = sagactx.saga_params::<Params>()?;
+    let opctx = OpContext::for_saga_action(&sagactx, &params.serialized_authn);
+    let osagactx = sagactx.user_data();
+    let dpd_client: Arc<dpd_client::Client> =


No need to clone, I think, the dpd-client methods we're calling take &self.

nexus/src/app/sagas/instance_create.rs

bnaecker · 2023-03-02T22:21:13Z

nexus/src/app/sagas/instance_create.rs

+                    .await
+            }
+        }
+        .map_err(|e| {


So I think the fact that each request can fail means these need to be separate saga actions. Look at the NIC and disk-creation for examples. We basically create N nodes, for N NICs. Each of those creates (or destroys) exactly one item. Otherwise, the action is not really idempotent.

Thinking about this more, I'm not certain this is the case. If any of the requests fail, the saga is supposed to unwind, right? Even if the node is re-attempted, any requests that were successful would not be re-sent because of the logic above this that checks for existing nat entries, so unless I'm missing something, I think this is actually idempotent.

Persisting our chat discussion here:

I had a misunderstanding of Steno's behavior. Since we don't run undo actions for failed nodes, the logic in sic_remove_network_config() would never be called in the event of a partial failure, which is why the overall behavior would not be idempotent.

bnaecker · 2023-03-02T22:46:37Z

nexus/src/app/sagas/instance_create.rs

+        }
+    }
+
+    for entry in entries_to_remove {


Same thing here about idempotency. We need a different forward and undo action for each external IP address.

~~Do we actually need a separate saga node for each ip, or do we need to make sure that we attempt to delete the nat entry for every existing ip at least once?~~

Implicitly handled based on above discussion

bnaecker · 2023-03-02T22:47:07Z

nexus/src/app/sagas/instance_delete.rs

+
+    let mut entries_to_remove: Vec<ExternalIp> = vec![];
+
+    for entry in external_ips {


Same thing here about idempotency. We need a different saga action for each external IP I think.

Persisting convo from chat:

It doesn't seem that any of the other saga nodes in instance_delete.rs are using subsagas, so here we're going to ensure that we have attempted to nat entries for all external ips before checking errors.

docs/boundary-services-a-to-z.adoc

Update softnpu & dendrite dependencies Update commands in helper scripts to use `swadm` Update sled-agent config.toml to use softnpu override feature Update a-to-z docs to reflect new process

bnaecker

Tons of great work @internet-diglett! This has been a long slog, your persistence is admirable. I've got a few questions and small comments, but this generally seems pretty solid to me!

.github/buildomat/jobs/build-and-test-linux.sh

common/src/nexus_config.rs

sled-agent/src/services.rs

bnaecker · 2023-03-21T18:32:10Z

sled-hardware/src/illumos/mod.rs

+                // TODO: correctness
+                // I'm not sure whether or not we should be treating softnpu
+                // as a stub or treating it as a real ASIC here, so this
+                // might change.
+                ScrimletMode::Softnpu => HardwareView::new_stub_tofino(true),


This refers to how we look for the driver and devices files. This can't be a "real" Tofino, though we may want to add a variant to TofinoView for the SoftNPU-specific stuff.

sled-hardware/src/lib.rs

smf/sled-agent/non-gimlet/config.toml

tools/setup_path.sh

smklein

Thanks for taking this work on, really appreciate you doing it. Herculean effort, which rockets the effectiveness of the product forward.
I have a smattering of comments about minor changes before we check in, but when in doubt, pretend I added a "do what he said" comment under everything @bnaecker said.
Adding a stub implementation of dpd so cargo test can work without custom environment variables should be a pretty high priority for us to fix -- lmk if I can help get this work done!

smklein · 2023-02-28T21:14:57Z

sled-agent/src/config.rs

+    /// TODO: @internet-diglett
+    /// this is to preserve the old behavior of `scrimlet_override = false`,
+    /// but I haven't found where that logic has actually been leveraged...


In the old case, this was:

stub_scrimlet = Some(true) -> Please treat the sled as a scrimlet
stub_scrimlet = Some(false) -> Please DO NOT treat the sled as a scrimlet
stub_scrimlet = None -> Use whatever you automatically detect by scanning hardware

Basically, without this option, all "not-on-a-real-rack" deployments of Omicron look like Scrimlets.

Not sure if this case actually matters all too much, but that was the intent, at least.

smklein · 2023-03-21T19:08:36Z

.cargo/config

+# We need to enable this to access private repos
+[net]
+git-fetch-with-cli = true


Echoing @bnaecker 's comment here: https://github.com/oxidecomputer/omicron/pull/2089/files#r1143596296 , is this still necessary?

I don't think so. Removing it and seeing what happens in CI.

smklein · 2023-03-21T19:30:07Z

nexus/src/app/sagas/instance_create.rs

+
+    debug!(log, "checking for existing nat mapping for {target_ip:#?}");
+
+    // TODO: currently if we have this environment variable set, we want to


Can we file an issue to track this, and label these comments with TODO(issue number)?

smklein · 2023-03-21T19:33:41Z

README.adoc

+----
+SKIP_ASIC_CONFIG=1 cargo test
+----


I understand why this environment variable was added, and I empathize with the tradeoff being made here, but IMO the disposition is hostile to developers.

The property that "you should expect cargo build and cargo test to work" is pretty important - for example, a lot of effort was expended by @davepacheco to ensure that CockroachDB can be automatically started and used by our test suite, without requiring some environment variable setup.

You have comments down below with "TODOs" that we should be using a stub dpd, depending on how the platform was built, and I totally agree with that - we have support in our integration tests for launching an out-of-band CockroachDB for database testing, and we could totally do something similar for the stub dpd - it should be easy to have an in-Omicron implementation that can act as a fake.

That doesn't need to block this PR, fwiw. LMK if you want to pair on getting that stub in place.

Agreed on all points. I would like to work on getting that done ASAP.

smklein · 2023-03-21T19:35:03Z

common/src/nexus_config.rs

+/// Configuration for the dataplane daemon.
+#[derive(Clone, Debug, Deserialize, PartialEq, Serialize)]
+pub struct DpdConfig {
+    pub address: SocketAddr,


On a real system, where should this address come from? Is it on the underlay network? Would we expect it to have an entry in the internal DNS system?

On a real system there will be a dpd instance per scrimlet, each with a unique underlay address. DpdConfig is currently a single-valued member dpd_api in PackageConfig. This works for single-sidecar setups, but we'll need to evolve toward being able to handle multiple dpd instances.

I've noted this in the external networking plumbing I'm currently working on. I'd be happy to take this issue on in that work if we think that makes sense. There is additional infrastructure in the database in that branch that helps to identify the sidecar of interest through rack_id and switch_location when operations come through the API that require dpd interaction.

smklein · 2023-03-21T19:45:23Z

docs/boundary-services-a-to-z.adoc

+
+Follow the
+https://github.com/oxidecomputer/meta/blob/master/engineering/remote-access-preview-demo-setup.adoc#setting-up-the-cli[RAP document]
+to set up IPs, images, disks, instances etc. Things to pay particular attention
+to here are the following.


This is an internal document; IMO we should not be referencing it in public-facing docs (not necessarily for secrecy, but just because it won't make sense to anyone working with this codebase who isn't an Oxide employee).

smklein · 2023-03-21T19:49:56Z

illumos-utils/src/opte/illumos/mod.rs

-    // TODO-remove
-    //
-    // See https://github.com/oxidecomputer/omicron/issues/1337
-    //
-    // An additional part of the workaround to connect into instances. This is
-    // required to tell OPTE to actually act as a 1-1 NAT when an instance is
-    // provided with an external IP address, rather than do its normal job of
-    // encapsulating the traffic onto the underlay (such as for delivery to
-    // boundary services).
-    use_external_ip_workaround(&log, &xde_conf);


Hell yes, love to see this being deleted

illumos-utils/src/opte/illumos/mod.rs

smklein · 2023-03-21T20:02:50Z

tools/quickstart.sh

@@ -0,0 +1,87 @@
+#!/bin/bash


I'm kinda opposed to checking this script in - it's very environment-specific, may not make sense for all devs, but has a name which implies "you should run me first!"

I think this sorta thing makes sense for developers to have / use, but it's a bit of a landmine for someone to run without a lot of context about their environment

Would it make more sense for it to be something like example.sh instead of quickstart.sh? If not, I'm not opposed to removing it altogether.

I think an alternative would be to include the script or substantial portions of it in the documentation, such as how-to-run.adoc.

smklein · 2023-03-21T20:03:18Z

docs/boundary-services-a-to-z.adoc

+Once your host has been populated with images, you can use the script at
+`tools/quickstart.sh` to quickly create a VM and set up the `proxy-arp`. Please
+be sure to edit the `ip_pool_start` and `ip_pool_end` variables to match your
+desired address ranges.


See my comment below about this script; IMO we should not check it in.

bnaecker

Thanks for addressing all my questions, @internet-diglett! This all looks OK to me!

internet-diglett mentioned this pull request Jan 10, 2023

Implement the real networking stack #1828

Open

9 tasks

rcgoodfellow changed the title ~~Integrate softnpu~~ Integrate SoftNPP as virtual hardware Jan 31, 2023

internet-diglett changed the title ~~Integrate SoftNPP as virtual hardware~~ Integrate SoftNPU as virtual hardware Jan 31, 2023

morlandi7 added the networking Related to the networking. label Feb 7, 2023

rcgoodfellow reviewed Feb 10, 2023

View reviewed changes

tools/scrimlet/softnpu-init.sh Outdated Show resolved Hide resolved

rcgoodfellow reviewed Feb 10, 2023

View reviewed changes

docs/boundary-services-a-to-z.adoc Outdated Show resolved Hide resolved

internet-diglett force-pushed the integrate-softnpu branch from 06f4354 to 6b27f76 Compare February 15, 2023 21:47

internet-diglett marked this pull request as ready for review February 16, 2023 18:06

internet-diglett force-pushed the integrate-softnpu branch 2 times, most recently from 4bd4926 to 9fa6e34 Compare February 16, 2023 23:03

bnaecker reviewed Mar 2, 2023

View reviewed changes

smklein mentioned this pull request Mar 13, 2023

Create distinct configs for gimlet and non-gimlet devices #2518

Merged

rcgoodfellow assigned internet-diglett Mar 14, 2023

internet-diglett mentioned this pull request Mar 16, 2023

[Ignore] Test 2089 #2586

Closed

internet-diglett force-pushed the integrate-softnpu branch from e5563dc to b5eea6c Compare March 17, 2023 22:17

rcgoodfellow reviewed Mar 20, 2023

View reviewed changes

docs/boundary-services-a-to-z.adoc Outdated Show resolved Hide resolved

rcgoodfellow reviewed Mar 20, 2023

View reviewed changes

docs/boundary-services-a-to-z.adoc Outdated Show resolved Hide resolved

rcgoodfellow approved these changes Mar 20, 2023

View reviewed changes

rcgoodfellow and others added 6 commits March 20, 2023 21:06

Integrate softnpu from ry-working-softer-npu branch

901d314

Configure boundary services with dendrite

696d112

Update softnpu & dendrite dependencies Update commands in helper scripts to use `swadm` Update sled-agent config.toml to use softnpu override feature Update a-to-z docs to reflect new process

update dendrite

d5e780c

update softnpu machinery

443bd68

more dendrite updates

60bd11c

cleanup TODOs

01405e4

add local testing information to README

e013651

internet-diglett force-pushed the integrate-softnpu branch from 127635a to e013651 Compare March 20, 2023 21:37

smklein self-requested a review March 21, 2023 17:17

bnaecker reviewed Mar 21, 2023

View reviewed changes

smklein reviewed Mar 21, 2023

View reviewed changes

internet-diglett added 2 commits March 21, 2023 22:00

PR review fixes

308480f

feed clippy

4339551

bnaecker self-requested a review March 22, 2023 02:11

bnaecker approved these changes Mar 22, 2023

View reviewed changes

internet-diglett merged commit a796560 into main Mar 22, 2023

internet-diglett deleted the integrate-softnpu branch March 22, 2023 20:05

smklein mentioned this pull request Mar 23, 2023

local test failures trying to reach dpd #2649

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Integrate SoftNPU as virtual hardware #2089

Integrate SoftNPU as virtual hardware #2089

internet-diglett commented Dec 21, 2022 •

edited

Loading

rcgoodfellow commented Feb 11, 2023

rcgoodfellow commented Feb 11, 2023

rcgoodfellow commented Feb 24, 2023

rcgoodfellow commented Feb 24, 2023

rcgoodfellow commented Feb 26, 2023

rcgoodfellow commented Feb 27, 2023 •

edited

Loading

bnaecker left a comment

bnaecker Mar 2, 2023

bnaecker Mar 2, 2023

internet-diglett Mar 6, 2023

internet-diglett Mar 6, 2023

bnaecker Mar 2, 2023

internet-diglett Mar 6, 2023 •

edited

Loading

bnaecker Mar 2, 2023

internet-diglett Mar 6, 2023

bnaecker left a comment

bnaecker Mar 21, 2023

smklein left a comment

smklein Feb 28, 2023

smklein Mar 21, 2023

internet-diglett Mar 21, 2023

smklein Mar 21, 2023

internet-diglett Mar 21, 2023

smklein Mar 21, 2023

internet-diglett Mar 21, 2023

smklein Mar 21, 2023

rcgoodfellow Mar 21, 2023

smklein Mar 21, 2023

smklein Mar 21, 2023

smklein Mar 21, 2023

internet-diglett Mar 21, 2023

bnaecker Mar 22, 2023

smklein Mar 21, 2023

bnaecker left a comment


		let mut entries_to_remove: Vec<ExternalIp> = vec![];

		for entry in external_ips {


		debug!(log, "checking for existing nat mapping for {target_ip:#?}");

		// TODO: currently if we have this environment variable set, we want to

Integrate SoftNPU as virtual hardware #2089

Integrate SoftNPU as virtual hardware #2089

Conversation

internet-diglett commented Dec 21, 2022 • edited Loading

rcgoodfellow commented Feb 11, 2023

rcgoodfellow commented Feb 11, 2023

rcgoodfellow commented Feb 24, 2023

rcgoodfellow commented Feb 24, 2023

rcgoodfellow commented Feb 26, 2023

rcgoodfellow commented Feb 27, 2023 • edited Loading

bnaecker left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

internet-diglett Mar 6, 2023 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

bnaecker left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

smklein left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

bnaecker left a comment

Choose a reason for hiding this comment

internet-diglett commented Dec 21, 2022 •

edited

Loading

rcgoodfellow commented Feb 27, 2023 •

edited

Loading

internet-diglett Mar 6, 2023 •

edited

Loading