Update golang/go#11811
The increased default concurrency in Go 1.5 showed up a test flake in
the TestHostKeyCert test. Under load, when the client provided incorrect
data, both sides would race to tear down the connection, which would often
lead to the server side, running in its own goroutine to see an unexpected
EOF or connection reset.
Fix this flake (and the incorrect use of t.Fatalf) by passing the error back
to the main goroutine for inspection. This also lets us ignore the expected
error in the unsuccessful path
Change-Id: I5a95c6d240479e9d537f34177e5ca8023b1b08e9
Reviewed-on: https://go-review.googlesource.com/12916
Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org>
Fixesgolang/go#11882
If an error occurs during handshakeTransport.writePacket the lock may not be
released. Fix this by using defer rather than manually unlocking in all paths.
Change-Id: I0010284b4f7d99907c86b4c0e140ab6cf37b0441
Reviewed-on: https://go-review.googlesource.com/12888
Reviewed-by: Adam Langley <agl@golang.org>
Fixesgolang/go#9747
Move the example tests to an external test package so that they
must explicitly reference the ssh package. The side effect is the
examples now become easier to copy and paste.
Change-Id: Ibbddea42bc5a41d11ffdef5144d9884ef3ef603f
Reviewed-on: https://go-review.googlesource.com/3710
Reviewed-by: Andrew Gerrand <adg@golang.org>
This proposal effectively disables all the ssh/test tests when run with the -short flag supplied.
For developers and users of this package, there should be no change unless they are in the habbit of always supplying -short, which I belive is untrue.
For the CI dashboard the effect should be that these tests, which are really not portable enough to run reliably on all our various builders, are disabled.
LGTM=adg
R=adg, agl, hanwen
CC=golang-codereviews
https://golang.org/cl/125860043
Currently the ssh/terminal package cannot be compiled under gccgo. Even though gccgo may be running on linux, its syscall package is slightly different and does not contain these constants.
This proposal resolves the issue by declaring the two constants locally, as we've done for the *BSDs.
LGTM=hanwen, iant
R=hanwen, iant, gobot
CC=golang-codereviews
https://golang.org/cl/101670043
Transport should not be a ReadWriter. It can only write packets, i.e. no partial reads or writes. Furthermore, you can currently do ClientConn.Write() while the connection is live, which sends raw bytes over the connection. Doing so will confuse the transports because the data is not encrypted.
As a consequence, ClientConn and ServerConn stop being a net.Conn
Finally, ensure that {Server,Client}Conn implement LocalAddr and RemoteAddr methods that previously were exposed by an embedded net.Conn field.
R=hanwen
CC=golang-dev
https://golang.org/cl/16610043
Fixesgolang/go#5138.
Fixesgolang/go#4703.
This appears to pass my stress tests with and without the -race detector, but I'd like to see others hit it with their machines.
R=golang-dev, fullung, huin, kardianos, agl
CC=golang-dev
https://golang.org/cl/9929043
Followup CL for 9711043. The order that CombinedOutput returns data captured from stdout/stderr is not specified, so we have to test both variants.
Thanks to fullung for the bug report.
R=fullung, kr
CC=golang-dev
https://golang.org/cl/9921044
Fixesgolang/go#4285.
Adding a new cipher that is supported by the remote end, but not supported by our client causes that cipher to be considered a valid candidate. This fails later in setupKeys when there is no cipherModes configuration.
In summary, unsupported ciphers cannot be willed into existence by adding them to the client config. This change enforces this.
R=golang-dev, agl
CC=golang-dev
https://golang.org/cl/6780047
This proposal is an attempt to improve the state of functional testing in the ssh package. The previous functional tests required the user to give away some personal details, like their password and private key to run the tests, and so were probably not run as frequently as they should.
R=agl, gustav.paul, kardianos, fullung
CC=golang-dev
https://golang.org/cl/6601043
The check for a sensible packet length was removed a while ago
when the window size and channel packet size checks were moved
into channel.go. While the RFC suggests that any packet of size
less than uint32 -1 is valid, most implmentations limit the size
to a smaller value. OpenSSH chose 256kb, so that sounds like a
sensible default.
R=agl, huin, kardianos
CC=golang-dev
https://golang.org/cl/6490098
Fixesgolang/go#3810.
Fixes chanWriter Write after close behaviour bug.
Fixes serverChan writePacket after close bug.
Addresses final comments by agl on 6405064, plus various cleanups.
R=agl, kardianos, gustav.paul, fullung
CC=golang-dev
https://golang.org/cl/6479056
Fixesgolang/go#3989.
Tested for several hours on an 8 core ec2 instance with
random GOMAXPROC values.
Also, rolls server_test.go into session_test using the
existing dial() framework.
R=fullung, agl, kardianos
CC=golang-dev
https://golang.org/cl/6475063
Use a handler that does not attempt to send a status message
as the failing test closes the connection abruptly.
Also, check the err response on all shell.ReadLine operations.
R=agl, minux.ma, kardianos
CC=golang-dev
https://golang.org/cl/6487043
Fixesgolang/go#3810.
This change introduces an atomic boolean to guard the close
of the clientChan. Previously the client code was quite
lax with the ordering of the close messages and could allow
window adjustment or EOF messages to leak after Close had
been signaled.
Consolidating the changes to the serverChan will be handled
in a following CL.
R=agl, kardianos, gustav.paul
CC=golang-dev
https://golang.org/cl/6405064
This proposal moves the check for max packet into
channel.writePacket. Callers should be aware they cannot
pass a buffer larger than max packet. This is only a
concern to chanWriter.Write and appropriate guards are
already in place.
There was some max packet handling in transport.go but it was
incorrect. This has been removed.
This proposal also cleans up session_test.go.
R=gustav.paul, agl, fullung, huin
CC=golang-dev
https://golang.org/cl/6460075
RFC 4254 s5.2 is clear that a client must never send a data
packet larger than the value of maximum packet supplied by the
remote side during channel setup. The client was not honoring
this value, in fact it wasn't even recording it.
Thanks to Albert Strasheim for the bug report.
R=agl, fullung
CC=golang-dev
https://golang.org/cl/6448128
Fix a few resource leaks and prevent the test from
hanging if an error occurs reading from the remote
server.
R=agl, gustav.paul, kardianos
CC=golang-dev
https://golang.org/cl/6423065
This proposal removes the use of recover() to catch
invalid channel ids sent from the remote side. The
recover() unfortuntaly makes debugging harder as it
obscures other panic causes.
Another source of panic()s exists inside marshal.go,
which will be handled with in a later CL.
R=agl, gustav.paul
CC=golang-dev
https://golang.org/cl/6404046
The RFC doesn't prohibit zero sized window adjustments and
there is evidence of well known clients using them in the
wild.
R=agl, kardianos, gustav.paul
CC=golang-dev
https://golang.org/cl/6209082
I introduced this problem in change set 33 when I redefined
RejectionReason from a uint32 to an int. I have added a test
to verify the message can be marshaled correctly.
This was a bit hard to track down as the recover in clientConn
mainLoop would catch the panic and shutdown the connection. We
probably shouldn't be using panic inside marshal/unmarshal, at
least not without a way to let the user know why the recover
logic tripped.
R=agl, kardianos, gustav.paul
CC=golang-dev
https://golang.org/cl/6210063
This CL introduces a new struct, channel to hold common shared
functions.
* add a new channel struct, which is embeded in {client,server}Chan.
* move common methods from {client,server}Chan into channel.
* remove unneeded used of serverConn.lock in serverChan
(transport.writePacket has its own mutex).
* remove filteredConn, introduce conn.
R=agl, gustav.paul
CC=golang-dev
https://golang.org/cl/6128059
* Add support for RFC4254 section 4 global requests.
* Improve clientConn.Listen to process responses properly.
R=agl, gustav.paul
CC=golang-dev
https://golang.org/cl/6130050
Add support for server (remote) forwarded tcpip channels.
See RFC4254 Section 7.1
R=gustav.paul, jeff, agl, lieqiewang
CC=golang-dev
https://golang.org/cl/6038047
Hopefully fix build error under windows.
Binding to the wildcard is poor form for our darwin users
as it triggers the firewall popup. Dialing the wildcard
looks like it's implementation specific as well.
R=agl, kardianos
CC=golang-dev
https://golang.org/cl/6104046
Fixesgolang/go#3479.
Using a channel to model window size was a mistake. Unlike stdin and
stdout, which are streams of data, window size is an variable and
should be modeled as such.
R=golang-dev, agl, gustav.paul, kardianos, dvyukov
CC=golang-dev
https://golang.org/cl/5986053
Also, add support for hmac-sha1.
At the suggestion of AGL hmac-md5, and hmac-md5-96
support was not included.
Fixesgolang/go#3095.
R=golang-dev, agl, huin
CC=golang-dev
https://golang.org/cl/5696065