Where the spelling changed from British to
US norm (e.g., optimise -> optimize) it follows
the style in that file.
LGTM=adonovan
R=golang-codereviews, adonovan
CC=golang-codereviews
https://golang.org/cl/96980043
Update #3514
An io.Reader is permitted to return either (n, nil)
or (n, io.EOF) on EOF or other error.
The tls package previously always returned (n, nil) for a read
of size n if n bytes were available, not surfacing errors at
the same time.
Amazon's HTTPS frontends like to hang up on clients without
sending the appropriate HTTP headers. (In their defense,
they're allowed to hang up any time, but generally a server
hangs up after a bit of inactivity, not immediately.) In any
case, the Go HTTP client tries to re-use connections by
looking at whether the response headers say to keep the
connection open, and because the connection looks okay, under
heavy load it's possible we'll reuse it immediately, writing
the next request, just as the Transport's always-reading
goroutine returns from tls.Conn.Read and sees (0, io.EOF).
But because Amazon does send an AlertCloseNotify record before
it hangs up on us, and the tls package does its own internal
buffering (up to 1024 bytes) of pending data, we have the
AlertCloseNotify in an unread buffer when our Conn.Read (to
the HTTP Transport code) reads its final bit of data in the
HTTP response body.
This change makes that final Read return (n, io.EOF) when
an AlertCloseNotify record is buffered right after, if we'd
otherwise return (n, nil).
A dependent change in the HTTP code then notes whether a
client connection has seen an io.EOF and uses that as an
additional signal to not reuse a HTTPS connection. With both
changes, the majority of Amazon request failures go
away. Without either one, 10-20 goroutines hitting the S3 API
leads to such an error rate that empirically up to 5 retries
are needed to complete an API call.
LGTM=agl, rsc
R=agl, rsc
CC=golang-codereviews
https://golang.org/cl/76400046
Currently a write error will cause future reads to return that same error.
However, there may have been extra information from a peer pending on
the read direction that is now unavailable.
This change splits the single connErr into errors for the read, write and
handshake. (Splitting off the handshake error is needed because both read
and write paths check the handshake error.)
Fixes#7414.
LGTM=bradfitz, r
R=golang-codereviews, r, bradfitz
CC=golang-codereviews
https://golang.org/cl/69090044
While reviewing uses of the lower-level Client API in code, I found
that in many cases, code was using Client only because it needed a
timeout on the connection. DialWithDialer allows a timeout (and
other values) to be specified without resorting to the low-level API.
LGTM=r
R=golang-codereviews, r, bradfitz
CC=golang-codereviews
https://golang.org/cl/68920045
Currently an ECDHE handshake uses the client's curve preference. This
generally means that we use P-521. However, P-521's strength is
mismatched with the rest of the cipher suite in most cases and we have
a fast, constant-time implementation of P-256.
With this change, Go servers will use P-256 where the client supports
it although that can be overridden in the Config.
LGTM=bradfitz
R=bradfitz
CC=golang-codereviews
https://golang.org/cl/66060043
crypto/tls has two functions for creating a client connection: Dial,
which most users are expected to use, and Client, which is the
lower-level API.
Dial does what you expect: it gives you a secure connection to the host
that you specify and the majority of users of crypto/tls appear to work
fine with it.
Client gives more control but needs more care. Specifically, if it
wasn't given a server name in the tls.Config then it didn't check that
the server's certificates match any hostname - because it doesn't have
one to check against. It was assumed that users of the low-level API
call VerifyHostname on the certificate themselves if they didn't supply
a hostname.
A review of the uses of Client both within Google and in a couple of
external libraries has shown that nearly all of them got this wrong.
Thus, this change enforces that either a ServerName or
InsecureSkipVerify is given. This does not affect tls.Dial.
See discussion at https://groups.google.com/d/msg/golang-nuts/4vnt7NdLvVU/b1SJ4u0ikb0J.
Fixes#7342.
LGTM=bradfitz
R=golang-codereviews, bradfitz
CC=golang-codereviews
https://golang.org/cl/67010043
Users of the low-level, Client function are frequenctly missing the
fact that, unless they pass a ServerName to the TLS connection then it
cannot verify the certificates against any name.
This change makes it clear that at least one of InsecureSkipVerify and
ServerName should always be set.
LGTM=bradfitz
R=golang-codereviews, bradfitz
CC=golang-codereviews
https://golang.org/cl/65440043
Adam (agl@) had already done an initial review of this CL in a branch.
Added ClientSessionState to Config which now allows clients to keep state
required to resume a TLS session with a server. A client handshake will try
and use the SessionTicket/MasterSecret in this cached state if the server
acknowledged resumption.
We also added support to cache ClientSessionState object in Config that will
be looked up by server remote address during the handshake.
R=golang-codereviews, agl, rsc, agl, agl, bradfitz, mikioh.mikioh
CC=golang-codereviews
https://golang.org/cl/15680043
The renegotiation extension was introduced[1] due to an attack by Ray in
which a client's handshake was spliced into a connection that was
renegotiating, thus giving an attacker the ability to inject an
arbitary prefix into the connection.
Go has never supported renegotiation as a server and so this attack
doesn't apply. As a client, it's possible that at some point in the
future the population of servers will be sufficiently updated that
it'll be possible to reject connections where the server hasn't
demonstrated that it has been updated to address this problem.
We're not at that point yet, but it's good for Go servers to support
the extension so that it might be possible to do in the future.
[1] https://tools.ietf.org/search/rfc5746
R=golang-codereviews, mikioh.mikioh
CC=golang-codereviews
https://golang.org/cl/48580043
The practice of storing reference connections for testing has worked
reasonably well, but the large blocks of literal data in the .go files
is ugly and updating the tests is a real problem because their number
has grown.
This CL changes the way that reference tests work. It's now possible to
automatically update the tests and the test data is now stored in
testdata/. This should make it easier to implement changes that affect
all connections, like implementing the renegotiation extension.
R=golang-codereviews, r
CC=golang-codereviews
https://golang.org/cl/42060044
NSS (used in Firefox and Chrome) won't accept two certificates with the same
issuer and serial. But this causes problems with self-signed certificates
with a fixed serial number.
This change randomises the serial numbers in the certificates generated by
generate_cert.go.
R=golang-dev, r
CC=golang-dev
https://golang.org/cl/38290044
Despite SHA256 support being required for TLS 1.2 handshakes, some
servers are aborting handshakes that don't offer SHA1 support.
This change adds support for signing TLS 1.2 ServerKeyExchange messages
with SHA1. It does not add support for signing TLS 1.2 client
certificates with SHA1 as that would require the handshake to be
buffered.
Fixes#6618.
R=golang-dev, r
CC=golang-dev
https://golang.org/cl/15650043
Ticket 13740047 updated the documented TLS version to 1.2.
This also updates the RFC refered to.
R=golang-dev
CC=golang-dev, rsc
https://golang.org/cl/14029043
AES-GCM cipher suites are only defined for TLS 1.2, although there's
nothing really version specific about them. However, development
versions of NSS (meaning Firefox and Chrome) have an issue where
they'll advertise TLS 1.2-only cipher suites in a TLS 1.1 ClientHello
but then balk when the server selects one.
This change causes Go clients not to advertise TLS 1.2 cipher suites
unless TLS 1.2 is being used, and prevents servers from selecting them
unless TLS 1.2 has been negotiated.
https://code.google.com/p/chromium/issues/detail?id=297151https://bugzilla.mozilla.org/show_bug.cgi?id=919677
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/13573047
47ec7a68b1a2 added support for ECDSA ciphersuites but didn't alter the
cipher suite selection to take that into account. Thus Go servers could
try and select an ECDSA cipher suite while only having an RSA
certificate, leading to connection failures.
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/13239053
With TLS 1.2, when sending client certificates the code was omitting
the new (in TLS 1.2) signature and hash fields.
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/13413050
AES-GCM is the only current TLS ciphersuite that doesn't have
cryptographic weaknesses (RC4), nor major construction issues (CBC mode
ciphers) and has some deployment (i.e. not-CCM).
R=golang-dev, bradfitz
CC=golang-dev
https://golang.org/cl/13249044
Add support for ECDHE-ECDSA (RFC4492), which uses an ephemeral server
key pair to perform ECDH with ECDSA signatures. Like ECDHE-RSA,
ECDHE-ECDSA also provides PFS.
R=agl
CC=golang-dev
https://golang.org/cl/7006047
This does not include AES-GCM yet. Also, it assumes that the handshake and
certificate signature hash are always SHA-256, which is true of the ciphersuites
that we currently support.
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/10762044
Also use 2048-bit RSA keys as default in generate_cert.go,
as recommended by the NIST.
R=golang-dev, rsc, bradfitz
CC=golang-dev
https://golang.org/cl/10676043
TLS clients send ciphersuites in preference order (most prefereable
first). This change alters the order so that ECDHE comes before plain
RSA, and RC4 comes before AES (because of the Lucky13 attack).
This is unlikely to have much effect: as a server, the code uses the
client's ciphersuite order by default and, as a client, the non-Go
server probably imposes its order.
R=golang-dev, r, raggi, jsing
CC=golang-dev
https://golang.org/cl/10372045
The significant change between TLS 1.0 and 1.1 is the addition of an explicit IV in the case of CBC encrypted records. Support for TLS 1.1 is needed in order to support TLS 1.2.
R=golang-dev, bradfitz
CC=golang-dev
https://golang.org/cl/7880043
Currently we only check the leaf node's issuer against the list of
distinguished names in the server's CertificateRequest message. This
will fail if the client certiciate has more than one certificate in
the path and the leaf node issuer isn't in the list of distinguished
names, but the issuer's issuer was in the distinguished names.
R=agl, agl
CC=gobot, golang-dev
https://golang.org/cl/9795043
This isn't clearly a bug on Go's part, but it triggers a bug in Firefox
which means that crypto/tls and net/http cannot be wired up together
unless NextProtos includes "http/1.1". When net/http sets up the
tls.Config, it does this and so works fine. But anyone setting up the
tls.Config themselves will hit the Firefox bug.
Fixes#5445.
R=golang-dev, bradfitz, r
CC=golang-dev
https://golang.org/cl/9539045
A test added in b37d2fdcc4d9 didn't work with some values of GOMAXPROCS
because the defer statements were in the wrong order: the Pipe could be
closed before the TLS Client was.
R=golang-dev, bradfitz
CC=golang-dev
https://golang.org/cl/9187047
OpenSSL can be configured to send empty records in order to randomise
the CBC IV. This is an early version of 1/n-1 record splitting (that Go
does) and is quite reasonable, but it results in tls.Conn.Read
returning (0, nil).
This change ignores up to 100 consecutive, empty records to avoid
returning (0, nil) to callers.
Fixes 5309.
R=golang-dev, r, minux.ma
CC=golang-dev
https://golang.org/cl/8852044
Subject Alternative Names in X.509 certificates may include IP
addresses. This change adds support for marshaling, unmarshaling and
verifying this form of SAN.
It also causes IP addresses to only be checked against IP SANs,
rather than against hostnames as was previously the case. This
reflects RFC 6125.
Fixes#4658.
R=golang-dev, mikioh.mikioh, bradfitz
CC=golang-dev
https://golang.org/cl/7336046
Previously, Go TLS servers always took the client's preferences into
account when selecting a ciphersuite. This change adds the option of
using the server's preferences, which can be expressed by setting
tls.Config.CipherSuites.
This mirrors Apache's SSLHonorCipherOrder directive.
R=golang-dev, nightlyone, bradfitz, ality
CC=golang-dev
https://golang.org/cl/7163043
Add support for loading X.509 key pairs that consist of a certificate
with an EC public key and its corresponding EC private key.
R=agl
CC=golang-dev
https://golang.org/cl/6776043
The RFC doesn't actually have an opinion on whether this is a fatal or
warning level alert, but common practice suggests that it should be a
warning.
This involves rebasing most of the tests.
Fixes#3413.
R=golang-dev, shanemhansen, rsc
CC=golang-dev
https://golang.org/cl/6654050
I typoed the code and tried to parse all the way to the end of the
message. Therefore it fails when NPN is not the last extension in the
ServerHello.
Fixes#4088.
R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/6637052
Session resumption saves a round trip and removes the need to perform
the public-key operations of a TLS handshake when both the client and
server support it (which is true of Firefox and Chrome, at least).
R=golang-dev, bradfitz, rsc
CC=golang-dev
https://golang.org/cl/6555051
X509KeyPair wasn't really supposed to allow the certificate and
key to be in the same file, but it did work if you put the key
first. Since some HTTPS servers support loading keys and certs
like this, this change makes it work in either order.
Fixes#3986.
R=golang-dev, dave, rsc
CC=golang-dev
https://golang.org/cl/6499103
Fixes#3862.
There were many areas where conn.err was being accessed
outside the mutex. This proposal moves the err value to
an embedded struct to make it more obvious when the error
value is being accessed.
As there are no Benchmark tests in this package I cannot
feel confident of the impact of this additional locking,
although most will be uncontended.
R=dvyukov, agl
CC=golang-dev
https://golang.org/cl/6497070
