Skip to content
Open
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
130 changes: 130 additions & 0 deletions text/3907-mtls-registry-authentication.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,130 @@
- Feature Name: `mtls-registry-authentication`
- Start Date: 2026-01-16
- RFC PR: [rust-lang/rfcs#3907](https://github.com/rust-lang/rfcs/pull/3907)
- Rust Issue: [rust-lang/rust#0000](https://github.com/rust-lang/rust/issues/0000)

# Summary
[summary]: #summary

This is an RFC aimed at allowing Cargo to present client certificates when forming HTTP connections and support mutual TLS authentication with registries.

# Motivation
[motivation]: #motivation

Some organizations require client identity verification when interacting with privately hosted services. This can be achieved a number of ways, but is commonly done with certificates in a process called "mutual TLS" (mTLS).

Cargo does not currently support forwarding client certificate information when it configures its `libcurl` HTTP handle. This poses an issue for organizations that host private crate registries and perform client authentication via certificates, since there is no alternative way to forward these client provided certificates.

Authentication at the TLS level is different from the token-based methods for [Registry Authentication](https://doc.rust-lang.org/cargo/reference/registry-authentication.html) exposed by the [Credential Provider Protocol](https://doc.rust-lang.org/cargo/reference/credential-provider-protocol.html) since it takes place before the connection to the registry is established. It is not currently possible to write a [credential plugin](https://doc.rust-lang.org/cargo/reference/registry-authentication.html#credential-plugins) that enables this type of authentication with a registry, but an extension to that protocol would make this possible.

# Guide-level explanation
[guide-level-explanation]: #guide-level-explanation

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

What is our guidance on how registries should implement authentication? Should registries use MTLS? Should registries use tokens? Should registries use both? If the answer is, as it probably is, "It's complicated" then what are the factors that would drive that decision?

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

One operator data point: in my deployment they answer different questions and coexist. mTLS is the network-layer control: an identity-aware proxy (Teleport in my case, but Cloudflare Access or plain nginx ssl_verify_client are the same shape) decides whether the workload may reach the registry at all, before any HTTP request is made. The registry token remains the application-layer control: which repositories, read versus publish. Suggested guidance along those lines: mTLS authenticates the connection or workload, tokens authorize registry operations; a registry deployed behind an identity-aware proxy typically needs both, while a registry that terminates TLS itself and maps certificates to accounts may use mTLS alone. The deciding factor is whether the system checking the certificate and the system checking authorization are the same one.

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

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

Do identity-aware proxies forward the identity to the target application? If so this identity forwarding can take the place of the user provided token, right?

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

Yes, potentially. I think there are two separate axes here: how identity reaches the registry, and which operation or artifact is being authorized.

There are three identity-integration models:

  1. Admission only: the proxy authenticates the client, but the registry does not consume the forwarded identity.
  2. Direct assertion: the registry validates the proxy's signed identity assertion and maps its subject, roles, or traits to registry principals and permissions.
  3. Token exchange: an auth service validates the forwarded identity and issues a registry-native token with the appropriate audience and scopes, roughly the pattern standardized by RFC 8693.

A pattern I've used and seen across projects is to allow ordinary package downloads without application-layer credentials because the registry endpoint is already protected by an identity-aware access layer. Those reads are anonymous to the registry, but they are not anonymously reachable.

Publish, yank, owner-management, and administrative operations still require application-level authorization. That could come from a user-supplied registry token, a directly consumed proxy assertion, or a registry token derived from that assertion.

Some reads also require application authorization. For example, software with export-control or other data-classification restrictions might only be readable by a particular team. If multiple classifications exist behind one registry endpoint, the registry must consume a trusted identity and enforce repository- or artifact-level read policy; admission through the outer proxy alone is not granular enough.

Teleport supports identity forwarding by sending a signed Teleport-Jwt-Assertion on upstream requests, including the user's roles and traits. It can also rewrite Authorization to Bearer {{internal.jwt}}, although that overrides any client-supplied Authorization header. Teleport JWT docs and header-rewrite docs.

So yes, forwarded identity can replace the Cargo-provided token when the registry understands it. But it does not remove the need for application authorization: whether another credential is needed depends on the operation, the artifact's classification, and whether the registry can map the forwarded identity to sufficiently granular policy.


Credential providers will be able to provide client certificates and private keys to Cargo via new request and response messages.

Cargo will issue a tls-identity request when configuring an HTTP client for a registry, and the returned identity will be used for subsequent communication to the same registry (within the current Cargo session).

@cazlo cazlo Jul 11, 2026

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

"The same registry" can involve several HTTPS origins: the sparse index URL, the dl URL from the registry's config.json (often a CDN on a different host), and the API URL used for publish, yank, owner, and search. I think the RFC should define which origins are authorized to receive the identity,
because the default behavior of the current implementation surface would be broader than any of them: on current master (0a28f7930c), registry traffic goes through the shared http_async client, and every request handle sets follow_location(true), so libcurl follows redirects internally. A certificate configured on that handle is presented to every redirect hop, including cross-origin ones. libcurl's documented cross-host protections for Authorization headers do not cover TLS client certificates, which are selected at handshake time.

Note that Cargo's token behavior already faces this question: with auth-required, the Authorization header is sent to whatever origin the dl template resolves to. I think the identity design should be stricter than that precedent, since a certificate is presented at handshake time to every connection made by the handle rather than attached to individual requests.

There is a second isolation question in the same place: all easy handles share one curl multi and its connection cache. With two registries using different identities in one invocation, connection reuse and HTTP/2 multiplexing must never match a connection that was established under a different client certificate. curl has had this bug class twice: CVE-2021-22924 (connection reuse matched despite differing certificate-related config) and CVE-2022-27782 (TLS settings omitted from the reuse check entirely, fixed in 7.83.1). The second matters here because Cargo often links a system libcurl older than the fix, so it seems worth an explicit test rather than an assumption.

In the Go proof of concept linked above I scoped the certificate to one canonical HTTPS origin, used a distinct transport per identity so pooling cannot cross identity boundaries, and re-evaluated selection on every redirect hop. For Cargo the options seem to be: authorize only explicitly mapped origins, re-evaluate identity per hop by handling redirects in Cargo itself, or reject cross-origin redirects while an identity is loaded. Any of those can work; I mainly think the RFC text should pick one rather than leave it to the implementation.

View changes since the review


## TLS client identity request

* Sent by: Cargo
* Purpose: Get client certificates and private keys for HTTP communication

```json
{
// Protocol version
"v":2,
// Request kind: set TLS client identity
"kind":"tls-identity",
// Registry information (see https://doc.rust-lang.org/cargo/reference/credential-provider-protocol.html#registry-information)
"registry":{"index-url":"sparse+https://registry-url/index/"},
// Additional command-line args (optional)

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

What do these args represent? Where do they come from?

"args":[]
}
```

## TLS client identity response

* Sent by: credential provider
* Purpose: Set client certificates and private keys for HTTP communication

```json
{"Ok":{
// Response kind: this was a TLS client identity request
"kind":"tls-identity",
// Client certificate chain in PEM format, with escaped newlines (empty if unset)
"certificate":"-----BEGIN CERTIFICATE-----\n[Base64 encoded client certificate data]\n-----END CERTIFICATE-----",
// Private keys in PEM format, with escaped newlines (empty if unset)
"key":"-----BEGIN PRIVATE KEY-----\n[Base64 encoded private key data]\n-----END PRIVATE KEY-----"
}}
```

## Certificate and key formats

@cazlo cazlo Jul 11, 2026

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

In the Zulip thread the question came up whether HSM support via ssl_engine would lock Cargo into curl/OpenSSL. The same backend portability concern already applies to this response format on stock Windows builds today: Cargo's Windows libcurl is usually built with Schannel (see the comment in
src/cargo/util/network/http.rs), and libcurl documents that CURLOPT_SSLCERT_BLOB under Schannel requires PKCS#12, while CURLOPT_SSLKEY_BLOB is supported only by OpenSSL and wolfSSL. As drafted, the separate PEM certificate and key fields have no obvious implementation path on the default Windows toolchain.

I think the RFC should pick one of:
(a) Cargo converts the PEM pair in memory to whatever the backend needs (PEM to PKCS#12 for Schannel)
(b) an intentional, documented platform limitation with a clear fail-closed error, since silently continuing without client authentication would be the worst outcome
(c) a format field negotiated per backend behind the protocol.

Related: since a signing-oracle flow (the provider signs a digest and the key never leaves an HSM) is the likely future answer to the TPM/smartcard discussion above, it may be worth wording the response so that key is not structurally mandatory forever, for example "exactly one of key or a future signing capability", so protocol v2 can grow that without a breaking change.

View changes since the review


The `certificate` and `key` fields are expected to correspond to the same TLS client identity. If a credential provider is unable to supply a usable client identity, it may return empty fields.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

When should it be returning "empty fields" vs "Err":{"kind":"not-found"}?

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

One data point from how the token path handles this today: Cargo's provider chain treats url-not-supported and not-found as fall-through and tries the next configured provider, and only errors if no provider produced a result. If "no identity" is represented by empty fields in an Ok response, a multi-provider configuration behaves differently for identities than for tokens: the chain stops at the first provider even though it had nothing.

I'd suggest not-found for "this provider has no identity for this registry/origin", url-not-supported for "this provider doesn't handle this registry at all", and treating empty certificate/key fields in an Ok response as a protocol error. That keeps fallback semantics identical across request kinds, which also matters for mixed deployments where one provider serves tokens and a different one serves TLS identities.


The `certificate` field contains the client certificate chain in PEM format, with newlines escaped using `\n`. If multiple certificates are present, they are expected to be concatenated PEM blocks.

The `key` field contains the private key corresponding to the client certificate, in PEM format, with newlines escaped using `\n`.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

Every Cryptography expert I've ever talked to has strong opinions about PEM. The fact that it is widely used means that it's widely supported, but that doesn't necessarily mean we should start using it. I would want opinions from Rust Crypto before making this decision.


Encrypted private keys are not supported. Credential providers are responsible for decrypting user-provided material before returning it to Cargo.

# Reference-level explanation
[reference-level-explanation]: #reference-level-explanation

The currently used crate for `libcurl` exposes methods for setting these certificates and keys, and can be used to set these configuration options when HTTP handles are being configured. These methods are:
* `curl::easy::Easy::ssl_cert_blob`
* `curl::easy::Easy::ssl_key_blob`

These "easy" methods wrap well tested code in the curl source:
* https://github.com/curl/curl/blob/master/docs/libcurl/opts/CURLOPT_SSLKEY.md
* https://github.com/curl/curl/blob/master/docs/libcurl/opts/CURLOPT_SSLCERT.md

# Security Considerations

Cargo MUST treat all certificate and private key data returned by a credential provider as sensitive material.

Cargo MUST NOT persist tls-identity response data to disk.

All certificate and key material should be held in memory only for the lifetime required to configure the HTTP client.

Cargo MUST NOT log, print, or otherwise expose the contents of these blobs, including in debug or trace output.

Credential providers are responsible for securely sourcing and protecting private key material.

# Drawbacks
[drawbacks]: #drawbacks

This adds additional complexity to Cargo's HTTP configuration and could have impacts on where in the code HTTP handles are configured, and which handles are used for communication with different registries.

# Rationale and alternatives
[rationale-and-alternatives]: #rationale-and-alternatives

## Allow custom credential plugins to handle certificates

- Credential management is hard, and Cargo does not need to be directly involved in managing these certificates.

- There are near endlessly niche ways that a user may want to provide their client certificates, and this protocol extension will allow users to write custom plugins to handle their situation.

- This RFC does not introduce new trust boundaries beyond those already present for credential providers, which are treated as fully trusted by the user.

## Avoid backend lock-in

- Today Cargo is using `libcurl` for its backend HTTP client. There might be a future where a `rustls` based backend would be preferred. Nearly all TLS libraries support client certificates in some form, and this protocol extension gives Cargo ability to convert from the widely used PEM format to whatever may be needed in the future.

# Prior art
[prior-art]: #prior-art

Mutual TLS authentication is widely supported across TLS libraries, developer tools, and artifact distribution systems.

libcurl has supported client certificates via CURLOPT_SSLCERT and CURLOPT_SSLKEY since version 7.1 (released August 2000), and these options are commonly used by applications that require authenticated HTTPS connections. Other widely used TLS implementations, including OpenSSL, BoringSSL, NSS, and rustls, also provide first-class support for configuring client certificates and private keys for TLS connections.

Many developer tools and package managers support mutual TLS when interacting with private registries or artifact repositories. For example, Python package management tools such as Poetry and uv allow users to configure client certificates for authenticated registry access. Other ecosystems similarly support client certificate authentication, including pip, npm, Maven, and Gradle, where mutual TLS is commonly used in enterprise environments.

Private artifact repository systems and registry infrastructure, such as JFrog Artifactory, Sonatype Nexus, GitHub Enterprise, and GitLab, frequently support or encourage mutual TLS as an authentication mechanism for internal services. These systems are often deployed in environments with existing public key infrastructure, where TLS-level client authentication integrates naturally with organizational security policies.

Within Cargo itself, this RFC builds on existing design patterns established by the credential provider protocol. Cargo already delegates authentication concerns to external credential providers and avoids managing long-lived secrets directly. Extending this protocol to allow credential providers to supply TLS client identity material follows the same approach and enables mutual TLS support without introducing new secret management responsibilities into Cargo.

# Unresolved questions
[unresolved-questions]: #unresolved-questions

This RFC is intentionally limited to providing client certificate and private key material to Cargo. It does not address configuring additional certificate authority (CA) roots, interacting with platform trust stores, or performing certificate signing request (CSR)–based authentication flows (which would be needed to support hardware security modules). Future extensions to the credential provider protocol might allow credential providers to supply additional trust anchors or to participate in dynamic certificate issuance mechanisms, but these design decisions would likely be influenced by particular TLS backend choices.