An idea that came up in https://caddy.community/t/save-internally-issued-wildcard-certificate-in-consul/11740, this a simple module that might be useful for anyone who uses storage modules that aren't filesystem, to let them load certs/keys externally issued for use by Caddy.
Bit goofy, since we need to fetch the certmagic.Storage during provisioning, it needs a wrapping struct instead of just being an array like `load_files`.
Future work might involve adding Caddyfile support via a subdirective of the `tls` directive maybe?
The HTTP Caddyfile adapter can now configure the PKI app, and the acme_server directive can now be used to specify a custom CA used for issuing certificates. More customization options can follow later as needed.
If `tls <email>` is used, we should apply that to all applicable default issuers, not drop them. This refactoring applies implicit ACME issuer settings from the tls directive to all default ACME issuers, like ZeroSSL.
We also consolidate some annoying logic and improve config validity checks.
Ref: https://caddy.community/t/error-obtaining-certificate-after-caddy-restart/11335/8
This commits dds 3 separate, but very related features:
1. Automated server identity management
How do you know you're connecting to the server you think you are? How do you know the server connecting to you is the server instance you think it is? Mutually-authenticated TLS (mTLS) answers both of these questions. Using TLS to authenticate requires a public/private key pair (and the peer must trust the certificate you present to it).
Fortunately, Caddy is really good at managing certificates by now. We tap into that power to make it possible for Caddy to obtain and renew its own identity credentials, or in other words, a certificate that can be used for both server verification when clients connect to it, and client verification when it connects to other servers. Its associated private key is essentially its identity, and TLS takes care of possession proofs.
This configuration is simply a list of identifiers and an optional list of custom certificate issuers. Identifiers are things like IP addresses or DNS names that can be used to access the Caddy instance. The default issuers are ZeroSSL and Let's Encrypt, but these are public CAs, so they won't issue certs for private identifiers. Caddy will simply manage credentials for these, which other parts of Caddy can use, for example: remote administration or dynamic config loading (described below).
2. Remote administration over secure connection
This feature adds generic remote admin functionality that is safe to expose on a public interface.
- The "remote" (or "secure") endpoint is optional. It does not affect the standard/local/plaintext endpoint.
- It's the same as the [API endpoint on localhost:2019](https://caddyserver.com/docs/api), but over TLS.
- TLS cannot be disabled on this endpoint.
- TLS mutual auth is required, and cannot be disabled.
- The server's certificate _must_ be obtained and renewed via automated means, such as ACME. It cannot be manually loaded.
- The TLS server takes care of verifying the client.
- The admin handler takes care of application-layer permissions (methods and paths that each client is allowed to use).\
- Sensible defaults are still WIP.
- Config fields subject to change/renaming.
3. Dyanmic config loading at startup
Since this feature was planned in tandem with remote admin, and depends on its changes, I am combining them into one PR.
Dynamic config loading is where you tell Caddy how to load its config, and then it loads and runs that. First, it will load the config you give it (and persist that so it can be optionally resumed later). Then, it will try pulling its _actual_ config using the module you've specified (dynamically loaded configs are _not_ persisted to storage, since resuming them doesn't make sense).
This PR comes with a standard config loader module called `caddy.config_loaders.http`.
Caddyfile config for all of this can probably be added later.
COMMITS:
* admin: Secure socket for remote management
Functional, but still WIP.
Optional secure socket for the admin endpoint is designed
for remote management, i.e. to be exposed on a public
port. It enforces TLS mutual authentication which cannot
be disabled. The default port for this is :2021. The server
certificate cannot be specified manually, it MUST be
obtained from a certificate issuer (i.e. ACME).
More polish and sensible defaults are still in development.
Also cleaned up and consolidated the code related to
quitting the process.
* Happy lint
* Implement dynamic config loading; HTTP config loader module
This allows Caddy to load a dynamic config when it starts.
Dynamically-loaded configs are intentionally not persisted to storage.
Includes an implementation of the standard config loader, HTTPLoader.
Can be used to download configs over HTTP(S).
* Refactor and cleanup; prevent recursive config pulls
Identity management is now separated from remote administration.
There is no need to enable remote administration if all you want is identity
management, but you will need to configure identity management
if you want remote administration.
* Fix lint warnings
* Rename identities->identifiers for consistency
Allows user to disable OCSP stapling (including support in the Caddyfile via the ocsp_stapling global option) or overriding responder URLs. Useful in environments where responders are not reachable due to firewalls.
This changes the signature of UnmarshalGlobalFunc but this is probably OK since it's only used by this repo as far as we know.
We need this change in order to "remember" the previous value in case a global option appears more than once, which is now a possibility with the cert_issuer option since Caddy now supports multiple issuers in the order defined by the user.
Bonus: the issuer subdirective of tls now supports one-liner for "acme" when all you need to set is the directory:
issuer acme <dir>
This allows for finer-grained control when choosing alternate chains than
simply the previous/Certbot-esque behavior of "choose first chain that
contains an issuer's common name." This update allows you to sort by
length (if optimizing for efficiency on the wire) and also to select the
chain with a specific root CommonName.
* ci: Use golangci's github action for linting
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Fix most of the staticcheck lint errors
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Fix the prealloc lint errors
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Fix the misspell lint errors
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Fix the varcheck lint errors
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Fix the errcheck lint errors
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Fix the bodyclose lint errors
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Fix the deadcode lint errors
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Fix the unused lint errors
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Fix the gosec lint errors
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Fix the gosimple lint errors
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Fix the ineffassign lint errors
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Fix the staticcheck lint errors
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Revert the misspell change, use a neutral English
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Remove broken golangci-lint CI job
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Re-add errantly-removed weakrand initialization
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* don't break the loop and return
* Removing extra handling for null rootKey
* unignore RegisterModule/RegisterAdapter
Co-authored-by: Mohammed Al Sahaf <msaa1990@gmail.com>
* single-line log message
Co-authored-by: Matt Holt <mholt@users.noreply.github.com>
* Fix lint after a1808b0dbf209c615e438a496d257ce5e3acdce2 was merged
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Revert ticker change, ignore it instead
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Ignore some of the write errors
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Remove blank line
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Use lifetime
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* close immediately
Co-authored-by: Matt Holt <mholt@users.noreply.github.com>
* Preallocate configVals
Signed-off-by: Dave Henderson <dhenderson@gmail.com>
* Update modules/caddytls/distributedstek/distributedstek.go
Co-authored-by: Mohammed Al Sahaf <msaa1990@gmail.com>
Co-authored-by: Matt Holt <mholt@users.noreply.github.com>
* caddytls: Support multiple issuers
Defaults are Let's Encrypt and ZeroSSL.
There are probably bugs.
* Commit updated integration tests, d'oh
* Update go.mod
We have users that have site blocks like *.*.tld with on-demand TLS
enabled. While *.*.tld does not qualify for a publicly-trusted cert due
to its wildcards, On-Demand TLS does not actually obtain a cert with
those wildcards, since it uses the actual hostname on the handshake.
This improves on that logic, but I am still not 100% satisfied with the
result since I think we need to also check if another site block is more
specific, like foo.example.tld, which might not have on-demand TLS
enabled, and make sure an automation policy gets created before the
more general policy with on-demand...
Update internal issuer for compatibility -- yay simpler code!
The .1 version also fixes non-critical SAN extensions that caused trust
issues on several clients.
* caddytls: Add support for ZeroSSL; add Caddyfile support for issuers
Configuring issuers explicitly in a Caddyfile is not easily compatible
with existing ACME-specific parameters such as email or acme_ca which
infer the kind of issuer it creates (this is complicated now because
the ZeroSSL issuer wraps the ACME issuer)... oh well, we can revisit
that later if we need to.
New Caddyfile global option:
{
cert_issuer <name> ...
}
Or, alternatively, as a tls subdirective:
tls {
issuer <name> ...
}
For example, to use ZeroSSL with an API key:
{
cert_issuser zerossl API_KEY
}
For now, that still uses ZeroSSL's ACME endpoint; it fetches EAB
credentials for you. You can also provide the EAB credentials directly
just like any other ACME endpoint:
{
cert_issuer acme {
eab KEY_ID MAC_KEY
}
}
All these examples use the new global option (or tls subdirective). You
can still use traditional/existing options with ZeroSSL, since it's
just another ACME endpoint:
{
acme_ca https://acme.zerossl.com/v2/DV90
acme_eab KEY_ID MAC_KEY
}
That's all there is to it. You just can't mix-and-match acme_* options
with cert_issuer, because it becomes confusing/ambiguous/complicated to
merge the settings.
* Fix broken test
This test was asserting buggy behavior, oops - glad this branch both
discovers and fixes the bug at the same time!
* Fix broken test (post-merge)
* Update modules/caddytls/acmeissuer.go
Fix godoc comment
Co-authored-by: Francis Lavoie <lavofr@gmail.com>
* Add support for ZeroSSL's EAB-by-email endpoint
Also transform the ACMEIssuer into ZeroSSLIssuer implicitly if set to
the ZeroSSL endpoint without EAB (the ZeroSSLIssuer is needed to
generate EAB if not already provided); this is now possible with either
an API key or an email address.
* go.mod: Use latest certmagic, acmez, and x/net
* Wrap underlying logic rather than repeating it
Oops, duh
* Form-encode email info into request body for EAB endpoint
Co-authored-by: Francis Lavoie <lavofr@gmail.com>
When client certificate is enabled Caddy check only last certificate from
request. When this cert is not in list of trusted leaf certificates,
connection is rejected. According to RFC TLS1.x the sender's certificate
must come first in the list. Each following certificate must directly
certify the one preceding it.
This patch fix this problem - first certificate is checked instead of last.
This can lead to nicer, smaller JSON output for Caddyfiles like this:
a {
tls internal
}
b {
tls foo@bar.com
}
i.e. where the tls directive only configures automation policies, and
is merely meant to enable TLS on a server block (if it wasn't implied).
This helps keeps implicit config implicit.
Needs a little more testing to ensure it doesn't break anything
important.
* pki: Initial commit of embedded ACME server (#3021)
* reverseproxy: Support auto-managed TLS client certificates (#3021)
* A little cleanup after today's review session
- Create two default automation policies; if the TLS app is used in
isolation with the 'automate' certificate loader, it will now use
an internal issuer for internal-only names, and an ACME issuer for
all other names by default.
- If the HTTP Caddyfile adds an 'automate' loader, it now also adds an
automation policy for any names in that loader that do not qualify
for public certificates so that they will be issued internally. (It
might be nice if this wasn't necessary, but the alternative is to
either make auto-HTTPS logic way more complex by scanning the names in
the 'automate' loader, or to have an automation policy without an
issuer switch between default issuer based on the name being issued
a certificate - I think I like the latter option better, right now we
do something kind of like that but at a level above each individual
automation policies, we do that switch only when no automation
policies match, rather than when a policy without an issuer does
match.)
- Set the default LoggerName rather than a LoggerNames with an empty
host value, which is now taken literally rather than as a catch-all.
- hostsFromKeys, the function that gets a list of hosts from server
block keys, no longer returns an empty string in its resulting slice,
ever.
Certificate selection used to be a module, but this seems unnecessary,
especially since the built-in CustomSelectionPolicy allows quite complex
selection logic on a number of fields in certs. If we need to extend
that logic, we can, but I don't think there are SO many possibilities
that we need modules.
This update also allows certificate selection to choose between multiple
matching certs based on client compatibility and makes a number of other
improvements in the default cert selection logic, both here and in the
latest CertMagic.
The hardest part of this was the conn policy consolidation logic
(Caddyfile only, of course). We have to merge connection policies that
we can easily combine, because if two certs are manually loaded in a
Caddyfile site block, that produces two connection policies, and each
cert is tagged with a different tag, meaning only the first would ever
be selected. So given the same matchers, we can merge the two, but this
required improving the Tag selection logic to support multiple tags to
choose from, hence "tags" changed to "any_tag" or "all_tags" (but we
use any_tag in our Caddyfile logic).
Combining conn policies with conflicting settings is impossible, so
that should return an error if two policies with the exact same matchers
have non-empty settings that are not the same (the one exception being
any_tag which we can merge because the logic for them is to OR them).
It was a bit complicated. It seems to work in numerous tests I've
conducted, but we'll see how it pans out in the release candidates.
* tls: Support placeholders in key_type
* caddytls: Simplify placeholder support for ap.KeyType
Co-authored-by: Matthew Holt <mholt@users.noreply.github.com>
The comments in the code should explain the new logic thoroughly.
The basic problem for the issue was that we were overriding a catch-all
automation policy's explicitly-configured issuer with our own, for names
that we thought looked like public names. In other words, one could
configure an internal issuer for all names, but then our auto HTTPS
would create a new policy for public-looking names that uses the
default ACME issuer, because we assume public<==>ACME and
nonpublic<==>Internal, but that is not always the case. The new logic
still assumes nonpublic<==>Internal (on catch-all policies only), but
no longer assumes that public-looking names always use an ACME issuer.
Also fix a bug where HTTPPort and HTTPSPort from the HTTP app weren't
being carried through to ACME issuers properly. It required a bit of
refactoring.
When using the default automation policy specifically, ap.Issuer would
be nil, so we'd end up overwriting the ap.magic.Issuer's default value
(after New()) with nil; this instead sets Issuer on the template before
New() is called, and no overwriting is done.
* pki: Initial commit of PKI app (WIP) (see #2502 and #3021)
* pki: Ability to use root/intermediates, and sign with root
* pki: Fix benign misnamings left over from copy+paste
* pki: Only install root if not already trusted
* Make HTTPS port the default; all names use auto-HTTPS; bug fixes
* Fix build - what happened to our CI tests??
* Fix go.mod
* ability to specify that client cert must be present in SSL
* changed the clientauthtype to string and make room for the values supported by go as in caddy1
* renamed the config parameter according to review comments and added documentation on allowed values
* missed a reference
* Minor cleanup; docs enhancements
Co-authored-by: Matthew Holt <mholt@users.noreply.github.com>
This is a breaking change primarily in two areas:
- Storage paths for certificates have changed
- Slight changes to JSON config parameters
Huge improvements in this commit, to be detailed more in
the release notes.
The upcoming PKI app will be powered by Smallstep libraries.
We don't load the provider directly, because the lego provider types
aren't designed for JSON configuration and they are not implemented
as Caddy modules (there are some setup steps which a Provision call
would need to do, but they do not have Provision methods, they have
their own constructor functions that we have to wrap).
Instead of loading the challenge providers directly, the modules are
simple wrappers over the challenge providers, to facilitate the JSON
config structure and to provide a consistent experience. This also lets
us swap out the underlying challenge providers transparently if needed;
it acts as a layer of abstraction.
This ensure that if there are multiple certs that match a particular
ServerName or other parameter, then specifically the one the user
provided in the Caddyfile will be used.
It seems silly to have to add a single, empty TLS connection policy to
a server to enable TLS when it's only listening on the HTTPS port. We
now do this for the user as part of automatic HTTPS (thus, it can be
disabled / overridden).
See https://caddy.community/t/v2-catch-all-server-with-automatic-tls/6692/2?u=matt
This commit goes a long way toward making automated documentation of
Caddy config and Caddy modules possible. It's a broad, sweeping change,
but mostly internal. It allows us to automatically generate docs for all
Caddy modules (including future third-party ones) and make them viewable
on a web page; it also doubles as godoc comments.
As such, this commit makes significant progress in migrating the docs
from our temporary wiki page toward our new website which is still under
construction.
With this change, all host modules will use ctx.LoadModule() and pass in
both the struct pointer and the field name as a string. This allows the
reflect package to read the struct tag from that field so that it can
get the necessary information like the module namespace and the inline
key.
This has the nice side-effect of unifying the code and documentation. It
also simplifies module loading, and handles several variations on field
types for raw module fields (i.e. variations on json.RawMessage, such as
arrays and maps).
I also renamed ModuleInfo.Name -> ModuleInfo.ID, to make it clear that
the ID is the "full name" which includes both the module namespace and
the name. This clarity is helpful when describing module hierarchy.
As of this change, Caddy modules are no longer an experimental design.
I think the architecture is good enough to go forward.
* logging: Initial implementation
* logging: More encoder formats, better defaults
* logging: Fix repetition bug with FilterEncoder; add more presets
* logging: DiscardWriter; delete or no-op logs that discard their output
* logging: Add http.handlers.log module; enhance Replacer methods
The Replacer interface has new methods to customize how to handle empty
or unrecognized placeholders. Closes#2815.
* logging: Overhaul HTTP logging, fix bugs, improve filtering, etc.
* logging: General cleanup, begin transitioning to using new loggers
* Fixes after merge conflict
This migrates a feature that was previously reserved for enterprise
users, according to https://github.com/caddyserver/caddy/issues/2786.
Custom certificate selection policies allow advanced control over which
cert is selected when multiple qualify to satisfy a TLS handshake.
This migrates a feature that was previously reserved for enterprise
users, according to https://github.com/caddyserver/caddy/issues/2786.
TLS session ticket keys are sensitive, so they should be rotated on a
regular basis. Only Caddy does this by default. However, a cluster of
servers that rotate keys without synchronization will lose the benefits
of having sessions in the first place if the client is routed to a
different backend. This module coordinates STEK rotation in a fleet so
the same keys are used, and rotated, across the whole cluster. No other
server does this, but Twitter wrote about how they hacked together a
solution a few years ago:
https://blog.twitter.com/engineering/en_us/a/2013/forward-secrecy-at-twitter.html
This migrates a feature that was previously reserved for enterprise
users, according to https://github.com/caddyserver/caddy/issues/2786.
The PEM loader allows you to embed PEM files (certificates and keys)
directly into your config, rather than requiring them to be stored on
potentially insecure storage, which adds attack vectors. This is useful
in automated settings where sensitive key material is stored only in
memory.
Note that if the config is persisted to disk, that added benefit may go
away, but there will still be the benefit of having lesser dependence on
external files.
Making them pointers makes for cleaner JSON when adapting configs, if
the struct is empty now it will be omitted entirely.
The x/time/rate package was updated to support changing the burst, so
we've incorporated that here and removed a TODO.
