Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add fastly_acl hostcalls #438

Merged
merged 2 commits into from
Nov 4, 2024
Merged

Add fastly_acl hostcalls #438

merged 2 commits into from
Nov 4, 2024

Conversation

awilliams-fastly
Copy link
Contributor

@awilliams-fastly awilliams-fastly commented Oct 24, 2024

This adds support for two new hostcalls:

  • fastly_acl::open
  • fastly_acl::lookup

ACLs can be mocked via fastly.toml by adding one or more acls to local_server. Each acl entry is a reference to an external JSON file. Example:

[local_server]
acls.acl-v4 = "./acl-v4.json"
acls.acl-v6 = {file = "./acl-v6.json"}

The JSON file contains a list of one or more ACL entries, in the same format as can be used to modify ACLs via api.fastly.com. Example:

{
  "entries": [
    { "prefix": "1.2.3.0/24", "action": "BLOCK" },
    { "prefix": "192.168.0.0/16", "action": "BLOCK" },
    { "prefix": "23.23.23.23/32", "action": "ALLOW" },
    { "prefix": "1.2.3.4/32", "action": "ALLOW" },
    { "prefix": "127.0.0.0/8", "action": "ALLOW" }
  ]
}

The test-fixtures/src/bin/acl.rs file shows example guest usage.

Integration Tests

The fastly Rust crate containing support for ACLs has not yet been released. Because of that, the acl.rs integration test uses a configuration feature to conditionally compile, with a default of not using the fastly::acl feature:

// Temporary until fastly SDK is released which
// includes the fastly::acl module.
#[cfg(feature = "acl_hostcalls")]

To test against a local, unreleased crate:

  • Modify test-fixtures/Cargo.toml to point to the appropriate local crates
  • Compile with acl_hostcalls feature: (cd test-fixtures && cargo build -F acl_hostcalls --target=wasm32-wasi)

Once a new version of the crate has been published, the conditional can be removed. I have tested against the internal/unreleased crate and verified that it passes.

@awilliams-fastly awilliams-fastly marked this pull request as ready for review October 28, 2024 20:15
@awilliams-fastly awilliams-fastly changed the title Add fastly_acl hostcalls Add fastly_acl hostcalls Oct 28, 2024
fn main() -> Result<(), Error> {
// Temporary until fastly SDK is released which
// includes the fastly::acl module.
#[cfg(feature = "acl_hostcalls")]
Copy link
Contributor

@ulyssa ulyssa Oct 30, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If I'm following right, is the idea here to:

  • Merge this PR
  • Publish Viceroy
  • Publish Rust SDK with ACLs support
  • Do follow-up PR to update SDK version dependency and remove the acl_hostcalls feature guard

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, that's also how I was thinking the sequencing would be.

This feature guard can be removed once the new fastly crate is published (the order of the previous steps shouldn't matter).

#[derive(Clone, Debug, Default)]
pub struct AclConfig(pub(crate) acl::Acls);

mod deserialization {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It seems fine, but I'm curious: is there a reason to put this into its own deserialization module? And would it be worth putting it into its own lib/src/config/acl/deserialization.rs?

Copy link
Contributor Author

@awilliams-fastly awilliams-fastly Oct 31, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I choose this structure to match other similar sections (e.g backends, device_detection, dictionaries, ...). So not a very compelling reason.

I'll change it to whichever organization you think best.

ulyssa
ulyssa previously approved these changes Oct 30, 2024
Copy link
Contributor

@ulyssa ulyssa left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this looks good! I just have two minor questions.

mhp
mhp previously approved these changes Oct 31, 2024
Copy link
Contributor

@mhp mhp left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for doing this work. I've left a couple of comments regarding the json format, but I think I just convinced myself it was probably fine. I've also left a note to the effect that the goal here is to match production behaviour on a functional basis, not on a performance basis.

@@ -0,0 +1,8 @@
{
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I really like the fact that the file formats are close to the format used to send updates to the API. How do you feel about adding the "op" field too, so that the formats are the same?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, good idea. I've made that change in c18e3b6

{ "op": "update", "prefix": "192.168.0.0/16", "action": "BLOCK" },
{ "op": "create", "prefix": "23.23.23.23/32", "action": "ALLOW" },
{ "op": "update", "prefix": "1.2.3.4/32", "action": "ALLOW" },
{ "op": "update", "prefix": "1.2.3.4/8", "action": "ALLOW" }
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It looks like this testcase includes the "op" field, which is nice. Is it just ignored then? That's probably fine then!

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, it's ignored (along with any other extraneous fields).

Thanks for pointing out the inconsistency with this test case and the integration test data. In c18e3b6, I've corrected that along with adding better comments to better document the JSON structure.

/// over the acls entries.
///
/// If the IP matches multiple ACL entries, then:
/// - The most specific match is returned (longest mask),
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This looks like it gives the same results as we'd expect to see in production. Obviously the performance characteristics will be different, but I don't believe that it is a goal of Viceroy to duplicate that.

@awilliams-fastly awilliams-fastly dismissed stale reviews from mhp and ulyssa via c18e3b6 October 31, 2024 15:28
@awilliams-fastly awilliams-fastly merged commit eeac2f4 into main Nov 4, 2024
15 checks passed
@awilliams-fastly awilliams-fastly deleted the awilliams/acls branch November 4, 2024 23:01
awilliams-fastly added a commit that referenced this pull request Dec 5, 2024
This updates the test fixtures to use the newly released `0.11.1` Fastly crate.

When `fastly::acl` hostcalls were added in #438, the corresponding SDK support
was not yet released. So #438 added tests using the still-unreleased-SDK, but
put that code behind a compile-time gate ([here](#438 (comment))).

This change removes that compile-time gate in
`test-fixtures/src/bin/acl.rs`, effectively changing these tests from a
no-op to running the desired test cases.

A few other text-fixtures stopped compiling with the updated SDK, which
I've tried to address:

- `test-fixtures/src/bin/inspect.rs`
   cc/ @ulyssa

- `test-fixtures/src/bin/kv_store.rs`
   cc/ @joeshaw
awilliams-fastly added a commit that referenced this pull request Dec 5, 2024
This updates the test fixtures to use the newly released `0.11.1` Fastly crate.

When `fastly::acl` hostcalls were added in #438, the corresponding SDK support was not yet released. So #438 added tests using the still-unreleased-SDK, but put that code behind a compile-time gate ([here](#438 (comment))).

This change removes that compile-time gate in `test-fixtures/src/bin/acl.rs`, effectively changing these tests from a no-op to actually running the desired test cases.

A few other text-fixtures stopped compiling with the updated SDK, which I've tried to address as well:

- `test-fixtures/src/bin/inspect.rs`

- `test-fixtures/src/bin/kv_store.rs`

The following hostcalls were missing from the component/adapter, and so were added:
- `fastly_http_req#send_v3`
- `fastly_http_req#send_async_v2`
- `fastly_http_cache` (entire module)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants