Block when creating the renewal timer, rather than doing it in a
goroutine. This fixes an issue where startRenew and stopRenew are called
very closely together, and due to lock ordering, stopRenew may be called
before startRenew, resulting in the appearance that the renewal timer
has been stopped before it has actually been created.
This is only an issue in tests, as that is the only place stopRenew is
actually used. In particular this issue manifests in TestGetCertiifcate
sub-tests, where a httptest server reuses a port across two of the
sub-tests. In this case, the renewal calls end up creating dirty state
for the subsequent test, which can cause confusing behavior (such as
attempting to register an account twice.)
Another solution to this problem would be introducing a bool, protected
by renewalMu, which indicates if renewal has been halted, and to check
it in startRenew to check if stopRenew has already been called, which
would allow us to continue calling startRenew in a goroutine and relying
on renewalMu locking for ordering. That said I don't see a particularly
strong reason to call startRenew concurrently, so this seems like the
simplest solution for now.
Fixesgolang/go#52494
Change-Id: I95420d3fd877572a0b9e408d2f8cd353f6a4e80e
Reviewed-on: https://go-review.googlesource.com/c/crypto/+/433016
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Roland Shoemaker <roland@golang.org>
Reviewed-by: Bryan Mills <bcmills@google.com>
TestRenewFromCache and TestRenewFromCacheAlreadyRenewed had several
races and API misuses:
1. They called t.Fatalf from a goroutine other than the one invoking
the Test function, which is explicitly disallowed (see
https://pkg.go.dev/testing#T).
2. The test did not stop the renewal timers prior to restoring
test-hook functions, and the process of stopping the renewal timers
itself did not wait for in-flight calls to complete. That could
cause data races if one of the renewals failed and triggered a
retry with a short-enough randomized backoff.
(One such race was observed in
https://build.golang.org/log/1a19e22ad826bedeb5a939c6130f368f9979208a.)
3. The testDidRenewLoop hooks accessed the Manager.renewal field
without locking the Mutex guarding that field.
4. TestGetCertificate_failedAttempt set a testDidRemoveState hook, but
didn't wait for the timers referring to that hook to complete
before restoring it, causing races with other timers. I tried
pulling on that thread a bit, but couldn't untangle the numerous
untracked goroutines in the package. Instead, I have made a smaller
and more local change to copy the value of testDidRemoveState into
a local variable in the timer's closure.
Given the number of untracked goroutines in this package, it is likely
that races and/or deadlocks remain. Notably, so far I have been unable
to spot the actual cause of golang/go#51080.
For golang/go#51080
Change-Id: I7797f6ac34ef3c272f16ca805251dac3aa7f0009
Reviewed-on: https://go-review.googlesource.com/c/crypto/+/384594
Trust: Bryan Mills <bcmills@google.com>
Run-TryBot: Bryan Mills <bcmills@google.com>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
TryBot-Result: Gopher Robot <gobot@golang.org>
Let's Encrypt is revoking all certificates verified with TLS-ALPN-01
beofre January 26th due to a compliance issue. Detect them and force a
renewal.
Also, fix the tests which were not testing if expired certificates were
renewed anymore, as the test certificates were always invalid due to not
having SANs.
Change-Id: If9d0632b2edfe0b7fb70f6cfd7e65e46e2d047dc
Reviewed-on: https://go-review.googlesource.com/c/crypto/+/381114
Trust: Filippo Valsorda <filippo@golang.org>
Run-TryBot: Filippo Valsorda <filippo@golang.org>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: Roland Shoemaker <roland@golang.org>
Previously, the o.AuthzURLs slice was sometimes used from the call to
client.WaitOrder at the bottom of the for loop.
By that point, o may be nil if client.WaitOrder returned an error,
which would cause a nil pointer dereference panic inside the deferred
function call. If client.WaitOrder did not return an error, then the
call to deactivatePendingAuthz would use its AuthzURLs slice instead
of the one from client.AuthorizeOrder.
Fixesgolang/go#35225
Updates golang/go#21081
Change-Id: I7db055ee1149871b6e5d34a8618526899c68f827
Reviewed-on: https://go-review.googlesource.com/c/crypto/+/203919
Reviewed-by: Alex Vaghin <ddos@google.com>
The Manager now uses RFC 8555 implementation of Let's Encrypt by default.
Existing users need not do any manual upgrades. If you vendor
acme/autocert, it is enough to just rebuild your binaries at this CL.
If there's an account key stored in Manager's cache which has been used
with an earlier Let's Encrypt implementation (aka v1 or draft-02),
it will be automatically re-registered with the new endpoint.
One notable change is the CAServer from internal/acmetest was amended
to simulate a CA implementing RFC 8555, replacing draft-02.
Support for both RFC and draft-02 seemed too complicated and not worth
the benefits: the old pre-RFC bits will be removed from both acme and
acme/autocert packages at some point.
Fixesgolang/go#21081
Change-Id: Id530758ac612b1c20f9df51c4d10f770e5f41ecf
Reviewed-on: https://go-review.googlesource.com/c/crypto/+/199520
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
According to the [RFC 5280, section 4.2.1.6](https://tools.ietf.org/html/rfc5280#section-4.2.1.6):
> ...
> When the subjectAltName extension contains a domain name system
> label, the domain name MUST be stored in the dNSName (an IA5String).
> The name MUST be in the "preferred name syntax", as specified by
> Section 3.5 of [RFC1034] and as modified by Section 2.1 of
> [RFC1123]. Note that while uppercase and lowercase letters are
> allowed in domain names, no significance is attached to the case.
> ...
And the [RFC 1034, section 3.5](https://tools.ietf.org/html/rfc1034#section-3.5):
> ...
> Note that while upper and lower case letters are allowed in domain
> names, no significance is attached to the case. That is, two names with
> the same spelling but different case are to be treated as if identical.
> ...
We should return the same TLS certificate for both `example.com` and `EXAMPLE.COM`. But the `autocert.Manager.GetCertificate` treats the two as totally different, it signs and returns two different TLS certificates.
Moreover, now the `autocert.Manager.GetCertificate` and the `autocert.HostWhitelist` can only handle Punycode IDN. If the client sends a Unicode IDN to `autocert.Manager.GetCertificate` (cURL is doing this), the "Invalid character in DNS name" error will be triggered.
This PR corrects these problems by converting the host of the TLS certificate to be obtained to Punycode via `idna.Lookup.ToASCII`.
Change-Id: I993821b3a6ae532a53772e2db00524479ef111af
GitHub-Last-Rev: 6c12694574b6ac71a920da8a8bdb130c0b6e0272
GitHub-Pull-Request: golang/crypto#85
Reviewed-on: https://go-review.googlesource.com/c/crypto/+/171997
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
More informative error message enables HTTPS server configuration mistakes to be corrected quickly, since log files will now include the rejected hostname. If the hostname should be accepted, it can be added to the HostWhitelist Policy.
Fixesgolang/go#28345
Change-Id: I801c82f0d3b19bc34592c9cd0bce77f1b284d19d
GitHub-Last-Rev: 5dfe731d732d693f9bdd79c8a74efd30a2f56d0b
GitHub-Pull-Request: golang/crypto#63
Reviewed-on: https://go-review.googlesource.com/c/144337
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
The timeNow package variable doesn't work well here: since the renewal
functionality spawns goroutines that invoke timeNow, once a test has
caused such goroutines to exist, another test can't fiddle with it
without the race detector triggering.
Instead, have a private member of Manager that the tests can set if they
need.
Change-Id: Iaf1a68d8efb84c9c5e2804aeb9cc6b2d3f3fef43
Reviewed-on: https://go-review.googlesource.com/128655
Run-TryBot: Adam Langley <agl@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Because tls.Config now requires more fields to be set
in order for tls-alpn to work, Manager provides a new
TLSConfig method for easier setup.
This CL also adds a new internal package for end-to-end tests.
The package implements a simple ACME CA server.
Fixesgolang/go#25013Fixesgolang/go#25901
Updates golang/go#17251
Change-Id: I2687ea8d5c445ddafad5ea2cdd36cd4e7d10bc86
Reviewed-on: https://go-review.googlesource.com/125495
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
A var isn't needed and a const is what the upstream (Go 1.10+) version
it's copying is.
Change-Id: I335270be3b3d09ac3c22cf6fb889a74ac39b8f1d
Reviewed-on: https://go-review.googlesource.com/116379
Reviewed-by: Filippo Valsorda <filippo@golang.org>
GetCertificate has all the information it needs to know if a client
supports ECDSA in ClientHelloInfo. Deprecate and ignore ForceRSA, and
just obtain a RSA certificate on the fly when a client that doesn't
support ECDSA connects.
This changes the cache key format to have a "+rsa" suffix for RSA
certificates. The default (ForceRSA = false) cache key is unchanged,
so most DirCache instances will still be valid. Caches created with
ForceRSA set will be silently ignored and certificates reissued.
The cache keys for HTTP tokens and the account key are changed to be
guaranteed not to overlap with valid domain names as well.
Note that ECDSA support detection is more strict in following RFC 5246
than crypto/tls, which ignores signature_algorithms.
Fixesgolang/go#22066
Change-Id: I70227747b563d6849cb693f83a950d57040b3f39
Reviewed-on: https://go-review.googlesource.com/114501
Reviewed-by: Adam Langley <agl@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Fixes a bug introduced in golang.org/cl/100078 where incorrect
ACME client was used, causing nil pointer dereference.
The change also improves related tests,
removing code paths diverging in testing.
Fixesgolang/go#25581
Change-Id: I8c5531fcc5814a5a64f14911c0ad86c476a76d2f
Reviewed-on: https://go-review.googlesource.com/114817
Reviewed-by: Filippo Valsorda <filippo@golang.org>
We now keep track of pending authorization requests during verify() and
defer the asynchronous revocation of the ones that failed.
This should help avoid letsencrypt's "too many currently pending
authorizations" error.
Fixesgolang/go#23426
Change-Id: Ibffb10f59733962d45e43b67fc42a2ec7c5faf51
Reviewed-on: https://go-review.googlesource.com/100078
Run-TryBot: Filippo Valsorda <filippo@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Leo Antunes <costela@gmail.com>
Reviewed-by: Filippo Valsorda <filippo@golang.org>
The Manager now loops through known challenge types,
trying to fulfill one at a time until it succeeds or runs out
of supported challenges.
The provisioning of "http-01" challenges can be done
using the new Manager.HTTPHandler method.
It requires listening on unencrypted port 80.
Fixesgolang/go#21890
Change-Id: I55de9501f0069a9f460fedd8b5b0a09b94f9ef05
Reviewed-on: https://go-review.googlesource.com/87201
Run-TryBot: Alex Vaghin <ddos@google.com>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
If you don't use a cache you're just begging for Let's Encrypt
to ban you for a week due to duplicate certificate requests.
Don't encourage users to get banned.
Change-Id: Icf999c7f28817990bb45931d30692d4040c0ed71
Reviewed-on: https://go-review.googlesource.com/66550
Run-TryBot: Russ Cox <rsc@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Alex Vaghin <ddos@google.com>
This change makes the Manager try creating a certificate
again, after a previously unsuccessful attempt.
The implementation is based on a timer, to prevent hitting
an ACME CA with too high QPS when under a heavy load.
The timer is hardcoded to 1 minute.
Fixesgolang/go#17740.
Change-Id: I46a49201cf423be3360633a89209d7b2bccc1d76
Reviewed-on: https://go-review.googlesource.com/41694
Run-TryBot: Alex Vaghin <ddos@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
A cached cert data may be corrupted or simply contain an expired
certificate, which results in GetCertificate returning an error.
This change makes the Manager ignore those invalid and expired
cache entries, treating them as nonexistent.
Fixesgolang/go#20035.
Change-Id: I5345291ecb1aab1cf19671cf0a383135c7102038
Reviewed-on: https://go-review.googlesource.com/41690
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Without this, autocert will panic with an unhelpful nil pointer inside
the acme client.
Reorganized the test suite's ACME server stub creation, as I needed to
be able to stand up a test where GetCertificate was expected to fail.
Change-Id: Ie5e19c6e7766b4578c9b3c16789d7b27bd3be163
Reviewed-on: https://go-review.googlesource.com/40951
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
The standard library context package has been available since Go 1.7,
and the Go build dashboard (build.golang.org) only tests master and
the past two releases.
Also, the acme package makes no backwards compatibility guarantees,
not that I expect this to influence many people.
Change-Id: Ia1a294212e1e531f28f53cd954d4743a64611cab
Reviewed-on: https://go-review.googlesource.com/39272
Run-TryBot: Brad Fitzpatrick <bradfitz@golang.org>
Reviewed-by: Alex Vaghin <ddos@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Currently, autocert.Manager always generates EC-based certificates.
This change adds an optional field forcing the Manager to use RSA
instead.
An alternative idea, a "double" certificate, where the Manager
presents either RSA or EC certificate based on client's compatibility,
doesn't seem to be worth the implementation time given the constant
increase in Elliptic Curve cryptography.
Fixesgolang/go#17744
Change-Id: Idc68abfc698bcff4aad99715baefc06f8fae50ad
Reviewed-on: https://go-review.googlesource.com/34570
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
If no key is provided in the *acme.Client, one is generated, but it
is not cached. This means that every restart of the server process
will use a new account.
This change caches the account key, if one is generated, so that
restarts of the server process use the same key.
Change-Id: I80f127b2cc79745a854b220b8918724ca228e87a
Reviewed-on: https://go-review.googlesource.com/28980
Run-TryBot: Alex Vaghin <ddos@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Alex Vaghin <ddos@google.com>
This change also gets the Manager closer to being able
to cleanup in short-lived HTTP servers running in a long-lived binary.
Change-Id: I49db36156896acc76d4757146c26b99e1665423b
Reviewed-on: https://go-review.googlesource.com/28491
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
A cert renewal loop is started when a certificate is fetched
from cache for the first time or a new one is created.
At most one renew loop is running per domain.
Closes#16851
Change-Id: I3d5821d8d76e9f2d9b551d0976ebc4cf91647092
Reviewed-on: https://go-review.googlesource.com/27611
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
It is quite often the case users need to poll an authorization until it
reaches a final state.
This change adds a handy method to do exactly that,
as well as updating autocert.Manager to use the new method.
Change-Id: I4d25e4ee751731815b77980caab6da98d3440b5d
Reviewed-on: https://go-review.googlesource.com/27431
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Sanjay came up with this idea of a more flexible way
to place restrictions on the Manager using a HostPolicy hook
instead of the static DNSNames field.
HostPolicy allows for user-made custom policies,
as well as makes it possible to change the set of host names
dynamically, without restarting the Manager.
Change-Id: Ib7c6b047469edc6856b59c5e8365690e66f2a3a4
Reviewed-on: https://go-review.googlesource.com/27251
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Package autocert provides automatic access to certificates
from Let's Encrypt and any other ACME-based CA.
It is heavily based on the ideas from rsc.io/letsencrypt.
Change-Id: I62021452a918cd49093162f3d6c74e9d7f452fb8
Reviewed-on: https://go-review.googlesource.com/23970
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Roland Shoemaker
Ben Burkert
Bryan C. Mills
Filippo Valsorda
Trevor Dixon
Dmitri Shuralyov
Alex Vaghin
Aofei Sheng
David Jones
Adam Langley
Alex Vaghin
Leo Antunes
Ingo Gottwald
Kevin Burke
Russ Cox
Ross Light
Jonathan Rudenberg