feat(geoblock): let callers choose how the client IP is resolved - #37
Open
geekbrother wants to merge 2 commits into
Open
geekbrother wants to merge 2 commits into
geekbrother wants to merge 2 commits into
Conversation
`GeoBlockService::call` resolved the address through `InsecureClientIp`, which reads the leftmost `X-Forwarded-For` entry and then falls through `Forwarded`, `X-Real-Ip`, `Fly-Client-IP`, `True-Client-IP`, `CF-Connecting-IP` and `CloudFront-Viewer-Address` before reaching the connection peer. Every one of those is set by the caller, and a proxy in front of the service appends to `X-Forwarded-For` rather than replacing it, so the leftmost entry stays whatever the caller sent. A visitor in a blocked country could name an address in an unblocked one and pick its own geo answer. The extractor was not reachable from outside the crate, so a deployment that did know the real address — stamped into a header by a gateway or load balancer it controls — had no way to tell the layer to use it. Make the resolution a parameter. `ClientIpExtractor` wraps a closure over the request's headers and extensions, and `with_client_ip_extractor` on both the layer and the service takes one. `new` keeps its signature and its behaviour, now spelled `ClientIpExtractor::insecure_from_forwarding_headers`, so existing callers build and run unchanged and adopting the trusted path stays a deliberate edit. The name and its documentation say what it trusts, which the old call site did not. `None` from an extractor is `UnableToExtractIPAddress`, so it reaches the configured `BlockingPolicy` like any other extraction failure rather than quietly allowing the request. Also add tower's `util` feature as a dev-dependency. The middleware tests drive the layer through `ServiceExt` and `ServiceBuilder::service_fn`, neither of which is available without it, so `cargo test -p geoip --features middleware` could not compile on its own before this — it only worked when another crate in a wider build happened to turn the feature on. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
`cargo clippy --workspace --all-features --all-targets -- -D warnings`, which is what CI runs, rejects the inline `Arc<dyn Fn(..) -> ..>` under `clippy::type_complexity`. Give it a name. Keep the comment above the spoofed-header request under the comment width so its formatting does not depend on which rustfmt wraps it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
geekbrother
force-pushed
the
feat/geoblock-client-ip-extractor
branch
from
September 21, 2026 15:45
ba5e79a to
bf89495
Compare
geekbrother
marked this pull request as ready for review
September 22, 2026 12:28
chris13524
approved these changes
Sep 23, 2026
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
Why
The geo-block resolved the caller's address from forwarding headers that the caller itself sets, so a visitor in a blocked country could name an address somewhere else and choose its own answer. The extractor was private to the crate, so a deployment that did know the real address stamped into a header by a gateway it controls had no way to say so.
That is pay-core's situation today: its rate limiter, telemetry and analytics all trust the gateway-stamped header, and the geo-block alone kept reading the caller's.
Decisions
Made the resolution a parameter instead of changing the default, so nothing using this crate breaks and moving to a trusted source stays a deliberate edit.
newkeeps its exact behaviour but is now named for what it trusts, which the old call site never said out loud. I rejected hard-coding a specific trusted header, because each deployment stamps a different one.Nonefrom an extractor is treated as an ordinary extraction failure so the existingBlockingPolicydecides, rather than adding a second policy axis.Worth looking at
The insecure default reads seven client-settable headers, not just
X-Forwarded-For, so a test pins that list, a dependency bump could otherwise change which headers decide a block without anyone noticing. I also added tower'sutilas a dev-dependency: the middleware tests referenceServiceExtandservice_fnand so could never compile standalone before, and it's worth a second opinion on whether that's the right fix or whether I'm papering over a workspace feature quirk. Not a breaking change, and consumers need a tagged release before they can adopt it.The Pay-Core sibling PR https://github.com/WalletConnect/pay-core/pull/1777