fix(grpc): sanitize invalid UTF-8 in sockaddr conversion

[arpitjain099]

1. Explain what the PR does

Fixes #5292.

getSockaddr was the one place left in the gRPC event-to-proto conversion that assigned raw map[string]string values straight into proto3 string fields without UTF-8 sanitization. A getsockname / sockaddr_t argument can carry non-UTF-8 bytes:

• sun_path of an abstract AF_UNIX socket begins with a NUL byte and can hold arbitrary binary bytes (uninitialized tail of the 108-byte sun_path buffer, abstract namespace names, etc.).
• sin_addr / sin6_addr can likewise contain raw bytes that are not valid UTF-8.

proto3 string fields must be valid UTF-8, so proto.Marshal rejected these values with string field contains invalid UTF-8. That error propagated out of StreamEvents, killing the stream and disconnecting clients (grpcurl and others). The reporter confirmed via bisection that simply having getsockname in the event list triggered the crash.

The earlier UTF-8 hardening (#4894, #4916) sanitized nearly every other string path in this file but missed getSockaddr. This change closes that gap by routing the three affected fields through the existing sanitizeStringForProtobuf helper, matching the convention already used everywhere else in event_data.go.

2. Explain how to test it

Unit test added in pkg/server/grpc/event_data_test.go (Test_getSockaddr_InvalidUTF8). It feeds AF_UNIX, AF_INET, and AF_INET6 sockaddr arguments whose address/path fields contain invalid UTF-8 bytes through getEventData, then calls proto.Marshal on the resulting pb.Event.

Before this change the marshal fails:

proto: ...: string field contains invalid UTF-8

After it, marshal succeeds, the sanitized fields are valid UTF-8, and a valid sun_path like /tmp/socket is returned byte-for-byte unchanged.

make test-unit PKG=pkg/server/grpc TEST=Test_getSockaddr_InvalidUTF8

3. Other comments

The sanitization is lossy only for already-malformed input: invalid byte sequences are dropped (the same behavior sanitizeStringForProtobuf already applies to every other string field), while valid UTF-8 is preserved exactly. An alternative would be to expose the raw bytes through a dedicated bytes field on SockAddr, but that is an API/proto change; string sanitization keeps the wire format and existing consumers unchanged and is consistent with the rest of this file.

[geyslan]

@arpitjain099 thank you for contributing. @trvll do you mind reviewing this? Cheers.