This change will simplify follow-up work done as part of
https://github.com/ooni/probe/issues/1803#issuecomment-957323297 to
implement a comprehensive self-censoring solution.
While there, rename the "proxy" action to "pass" because what we
are effectively doing is passing traffic to the network (that's a
minor change but it seems a better analogy).
Reducing the errors is not done in a perfect way.
We have documented the most striking differences inside
https://github.com/ooni/probe/issues/1707#issuecomment-942283746 and
some attempts to improve the situation further inside
https://github.com/ooni/probe/issues/1707#issuecomment-942341255.
A better strategy for the future would be to introduce more
specific timeout errors, such as dns_timeout_error, etc.
More testing may be needed to further validate and compare the
old and the new TH, but this requires Jafar improvements to
more precisely simulate more complex censorship.
This is the most immediate fix to the issue described by
https://github.com/ooni/probe/issues/1792.
So, the logic was actually miss the increment, which
would have been noticed with proper unit testing.
Anyway, I am not sure why the loop ensues in the first
time. By looking at the headers, it seems we're passing
the headers correctly.
So, even though this fix interrupts the loop, it still
remains the question of whether the loop is legit or
whether we're missing extra logic to properly redirect.
This diff adds the prototype websteps implementation that used
to live at https://github.com/ooni/probe-cli/pull/506.
The code is reasonably good already and it's pointing to a roaming
test helper that I've properly configured.
You can run websteps with:
```
./miniooni -n websteps
```
This will go over the test list for your country.
At this stage the mechanics of the experiment is set, but we
still need to have a conversation on the following topics:
1. whether we're okay with reusing the data format used by other
OONI experiments, or we would like to use a more compact data
format (which may either be a more compact JSON or we can choose
to always submit compressed measurements for websteps);
2. the extent to which we would like to keep the measurement as
a collection of "the experiment saw this" and "the test helper
saw that" and let the pipeline choose an overall score: this is
clearly an option, but there is also the opposite option to
build a summary of the measurement on the probe.
Compared to the previous prototype of websteps, the main
architectural change we have here is that we are following
the point of view of the probe and the test helper is
much more dumb. Basically, the probe will choose which
redirection to follow and ask the test helper every time
it discovers a new URL to measure it w/o redirections.
Reference issue: https://github.com/ooni/probe/issues/1733
This commit introduce a measurement library that consists of
refactored code from earlier websteps experiments.
I am not going to add tests for the time being, because this library
is still a bit in flux, as we finalize websteps.
I will soon though commit documentation explaining in detail how
to use it, which currrently is at https://github.com/ooni/probe-cli/pull/506
and adds a new directory to internal/tutorial.
The core idea of this measurement library is to allow two
measurement modes:
1. tracing, which is what we're currently doing now, and the
tutorial shows how we can rewrite the measurement part of web
connectivity with measurex using less code. Under a tracing
approach, we construct a normal http.Client that however has
tracing configured, we gather events for resolve, connect, TLS
handshake, QUIC handshake, HTTP round trip, etc. and then we
try to make sense of what happened from the events stream;
2. step-by-step, which is what websteps does, and basically
means that after each operation you immediately write into
a Measurement structure its results and immediately draw the
conclusions on what seems odd (which later may become an
anomaly if we see what the test helper measured).
This library is also such that it produces a data format
compatible with the current OONI spec.
This work is part of https://github.com/ooni/probe/issues/1733.
This is required to implement websteps, which is currently tracked
by https://github.com/ooni/probe/issues/1733.
We introduce the concept of async runner. An async runner will
post measurements on a channel until it is done. When it is done,
it will close the channel to notify the reader about that.
This change causes sync experiments now to strictly return either
a non-nil measurement or a non-nil error.
While this is a pretty much obvious situation in golang, we had
some parts of the codebase that were not robust to this assumption
and attempted to submit a measurement after the measure call
returned an error.
Luckily, we had enough tests to catch this change in our assumption
and this is why there are extra docs and tests changes.
At the moment ooapi is not used. It will eventually be used since
it's a better way of accessing the OONI backend API.
To fix these tests, we need to fix the swagger emitted by the
backend API, which is not a priority at the moment, since we are
working instead to integrate websteps in miniooni.
Issue https://github.com/ooni/probe/issues/1790 tracks the work
required to re-enabled the tests I'm skipping with this diff.
This work is part of https://github.com/ooni/probe/issues/1733.
* feat: run ~always netxlite integration tests
This diff ensures that we check on windows, linux, macos that our
fundamental networking library (netxlite) works.
We combine unit and integration tests.
This work is part of https://github.com/ooni/probe/issues/1733, where
I want to have more strong guarantees about the foundations.
* fix(filtering/tls_test.go): make portable on Windows
The trick here is to use the wrapped error so to normalize the
different errors messages we see on Windows.
* fix(netxlite/quic_test.go): make portable on windows
Rather than using the zero port, use the `x` port which fails
when the stdlib is parsing the address.
The zero port seems to work on Windows while it does not on Unix.
* fix(serialresolver_test.go): make error more timeout than before
This seems enough to convince Go on Windows about this error
being really a timeout timeouty timeouted thingie.
On Windows, GetAddrInfoW is a syscall and the Go resolver does
not attempt to map errors beyond WSA_HOST_NOT_FOUND, which becomes
"no such host", which we map to "dns_nxdomain_error".
See https://github.com/golang/go/blob/go1.17.1/src/net/lookup_windows.go#L16.
To map more GetAddrInfoW errors, thus, we need to enhance our
error classifier to have system specific errors.
Then, we need to filter for the WSA errors that are most likely
to pop up and map them to OONI failures. Those are three:
- WSANO_DATA which we have from our own UDP resolver as well
and which we can map to `dns_no_answer`
- WSANO_RECOVERY which we don't have but existed for MK so
we will use `dns_non_recoverable_failure`, which was an MK error
- WSATRY_AGAIN which likewise we map to the error that MK
used to emit, so `dns_temporary_failure`
This diff should address https://github.com/ooni/probe/issues/1467.
I need to run test on Windows and I just discovered that:
1. the `errno_unix.go` filename does not mean anything because
`unix` is not a valid platform, so we need a filename for
each platform that we care about;
2. on Windows we need to use WSA prefixed names;
3. `i/e/session_psiphon.go` was not building because of the
migration from `netxlite/iox` to `netxlite`.
This diff attempts to fix all three issues.
The reference issue is https://github.com/ooni/probe/issues/1733,
because I was working on such an issue.
When preparing a tutorial for netxlite, I figured it is easier
to tell people "hey, this is the package you should use for all
low-level networking stuff" rather than introducing people to
a set of packages working together where some piece of functionality
is here and some other piece is there.
Part of https://github.com/ooni/probe/issues/1591
I have recently seen a data race related our way of
mutating the outgoing request to set the host header.
Unfortunately, I've lost track of the race output,
because I rebooted my Linux box before saving it.
Though, after inspecting why and and where we're mutating
outgoing requets, I've found that:
1. we add the host header when logging to have it logged,
which is not a big deal since we already emit the URL
rather than just the URL path when logging a request, and
so we can safely zap this piece of code;
2. as a result, in measurements we may omit the host header
but again this is pretty much obvious from the URL itself
and so it should not be very important (nonetheless,
avoid surprises and keep the existing behavior);
3. when the User-Agent header is not set, we default to
a `miniooni/0.1.0-dev` user agent, which is probably not
very useful anyway, so we can actually remove it.
Part of https://github.com/ooni/probe/issues/1733 (this diff
has been extracted from https://github.com/ooni/probe-cli/pull/506).
While there, modernize the way in which we run tests to avoid
depending on the fake files scattered around the tree and to
use some well defined mock structures instead.
Part of https://github.com/ooni/probe/issues/1591
There are a bunch of packages where we don't really need to depend
on netx but we can use local definitions that describe what we are
expecting from data structures we receive in input. This diff
addresses one of such cases.
Part of https://github.com/ooni/probe/issues/1591
I discovered which transport were used by apitool and made sure he gets the same transports now. While there, I discovered an issue with ooni/oohttp that has been fixed with cba9b1ce5e.
Part of https://github.com/ooni/probe/issues/1591
* netxlite: improve docs, tests, and code quality
* better documentation
* more strict testing of dialer (especially make sure we
document the quirk in https://github.com/ooni/probe/issues/1779
and we have tests to guarantee we don't screw up here)
* introduce NewErrWrapper factory for creating errors so we
have confidence we are creating them correctly
Part of https://github.com/ooni/probe/issues/1591
Adapt other places where it was not using a logger to either choose
a reasonable logger or disable logging for backwards compat.
See https://github.com/ooni/probe/issues/1591
This simplifies serializing errors to `*string`. It did not
occur to me before. It seems quite a nice improvement.
Part of https://github.com/ooni/probe/issues/1591
They are now more readable. I'll do another pass and start
separating integration testing from unit testing.
I think we need to have some always on integration testing
for netxlite that runs on macOS, linux, and windows.
See https://github.com/ooni/probe/issues/1591
No real functional change. A few are needed and they will come
next. With this diff I just wanted to do cosmetic changes and
documentation changes, to ensure this package is okay.
See https://github.com/ooni/probe/issues/1591
We need still to add similar wrappers to internal/netxlite but we
will adopt a saner approach to error wrapping this time.
See https://github.com/ooni/probe/issues/1591
For consistency and also because the SafeErrorWrapperBuilder seems
to be the building pattern than the original code assumed.
New code should not use it, but I'd rather keep legacy code consistent
formally and with its own original assumptions.
In particular, it matters that SafeErrorWrapperBuilder assigns the
most relevant operation that failed. We were not doing that when we
were manually creating a new ErrWrapper.
Part of https://github.com/ooni/probe/issues/1591
The legacy part for now is internal/errorsx. It will stay there until
I figure out whether it also needs some extra bug fixing.
The good part is now in internal/netxlite/errorsx and contains all the
logic for mapping errors. We need to further improve upon this logic
by writing more thorough integration tests for QUIC.
We also need to copy the various dialer, conn, etc adapters that set
errors. We will put them inside netxlite and we will generate errors in
a way that is less crazy with respect to the major operation. (The
idea is to always wrap, given that now we measure in an incremental way
and we don't measure every operation together.)
Part of https://github.com/ooni/probe/issues/1591
We will move the sane part of this package to i/netxlite/errorsx
and we will move the rest to i/e/legacy/errorsx.
What is the sane part? The sane part is error classifiers plus
the definition of ErrWrapper. The rest, including the rules
on how to decide whether an operation is major, are tricky and
we should consider them legacy and replace them with rules
that are more easy to understand and reason on.
Part of https://github.com/ooni/probe/issues/1591
No functional change, as it's clearly obvious from the output.
While there, also rename the generator for certifi. We are planning
on merging errorsx into netxlite. The first step is to give different
names to the code generating programs.
See https://github.com/ooni/probe/issues/1591
This is the last bit of functionality we need before rewriting a
chunk of websteps to use netxlite.
To rewrite most of it, we still need to move over:
1. dnstransport code
2. errorsx code
With both done, netxlite is a good library for websteps as well
as for most other operations we perform outside of the experiments.
Part of https://github.com/ooni/probe/issues/1591
With this change, we are now able to change more dependent code to simplify
the way in which we create and manage resolvers.
See https://github.com/ooni/probe/issues/1591
Like before, do not touch the rest of the tree. Rather create
compatibility types declared as legacy.
We will soon be able to close idle connections for an HTTP3
transport using any kind of resolvers more easily.
See https://github.com/ooni/probe/issues/1591
This quirk really saddens me. It's a piece of tech debt we're
carrying over from the original netx implementation.
We cannot remove it _until_ we have legacy netx code around.
The second best thing we can do is to clearly move this code in
a place where it's clear it's a quirk and write and use some extra
code that makes sure the quirk's assumptions are always met.
Sigh.
See https://github.com/ooni/probe/issues/1591
To make this happen, we need to take as argument a TLSDialer rather than
a TLSHandshaker. Then, we need to arrange the code so that we always
enforce a timeout for both TCP and TLS connections.
Because a TLSDialer can be constructed with a custom TLSConfig, we cover
also the case where the users wants to provide such a config.
While there, make sure we have better unit tests of the HTTP code.
See https://github.com/ooni/probe/issues/1591
While there reorganize mocks' tls implementation to use a single file
called tls.go (and tls_test.go) just like netxlite does.
While there write tests ensuring we always add timeouts when we are
making TCP connections (be them TLS or cleartext).
See https://github.com/ooni/probe/issues/1591
This basically adapts already existing code inside websteps to
instead be into the netxlite package, where it belongs.
In the process, abstract the TLSDialer but keep a reference to the
previous name to avoid refactoring existing code (just for now).
While there, notice that the right name is CloseIdleConnections (i.e.,
plural not singular) and change the name.
While there, since we abstracted TLSDialer to be an interface, create
suitable factories for making a TLSDialer type from a Dialer and a
TLSHandshaker.
See https://github.com/ooni/probe/issues/1591
We are proceeding with this plan of every major type being able to
close idle connections, which will simplify making DNS resolvers.
See https://github.com/ooni/probe/issues/1591.
Like we did before for the resolver, a dialer should propagate the
request to close idle connections to underlying types.
See https://github.com/ooni/probe/issues/1591
* fix(netxlite): make default resolver converge faster
Closes https://github.com/ooni/probe/issues/1726
* Update internal/netxlite/resolver.go
* fix(ndt7): adapt tests after previous change
Because now we're running the DNS resolution inside a goroutine
with a child context, the returned error string is different.
The previous error said we canceled the whole dialing operation,
while now we see directly that the context was canceled.
We would like to refactor the code so that a DoH resolver owns the
connections of its underlying HTTP client.
To do that, we need first to incorporate CloseIdleConnections
into the Resolver model. Then, we need to add the same function
to all netxlite types that wrap a Resolver type.
At the same time, we want the rest of the code for now to continue
with the simpler definition of a Resolver, now called ResolverLegacy.
We will eventually propagate this change to the rest of the tree
and simplify the way in which we manage Resolvers.
To make this possible, we introduce a new factory function that
adapts a ResolverLegacy to become a Resolver.
See https://github.com/ooni/probe/issues/1591.
## Description
This PR continues the refactoring of `netx` under the following principles:
1. do not break the rest of the tree and do not engage in extensive tree-wide refactoring yet
2. move under `netxlite` clearly related subpackages (e.g., `iox`, `netxmocks`)
3. move into `internal/netxlite/internal` stuff that is clearly private of `netxlite`
4. hide implementation details in `netxlite` pending new factories
5. refactor `tls` code in `netxlite` to clearly separate `crypto/tls` code from `utls` code
After each commit, I run `go test -short -race ./...` locally. Each individual commit explains what it does. I will squash, but this operation will preserve the original commit titles, so this will give further insight on each step.
## Commits
* refactor: rename netxmocks -> netxlite/mocks
Part of https://github.com/ooni/probe/issues/1591
* refactor: rename quicx -> netxlite/quicx
See https://github.com/ooni/probe/issues/1591
* refactor: rename iox -> netxlite/iox
Regenerate sources and make sure the tests pass.
See https://github.com/ooni/probe/issues/1591.
* refactor(iox): move MockableReader to netxlite/mocks
See https://github.com/ooni/probe/issues/1591
* refactor(netxlite): generator is an implementation detail
See https://github.com/ooni/probe/issues/1591
* refactor(netxlite): separate tls and utls code
See https://github.com/ooni/probe/issues/1591
* refactor(netxlite): hide most types but keep old names as legacy
With this change we avoid breaking the rest of the tree, but we start
hiding some implementation details a bit. Factories will follow.
See https://github.com/ooni/probe/issues/1591
The quic-go library does not support it anymore. So, let us be consistent
and remove any reference to h3-29 from our codebase.
Closes https://github.com/ooni/probe/issues/1740.
* fix: disable debianrepo build on master branch
This just mitigates https://github.com/ooni/probe/issues/1741 and does
not fully address it, but I'd rather avoid delving into this problem until
I open a release/v3.11.0 branch and have to really fix this issue.
* fix: only run coverage using go1.17
This is the version of Go with which we are going to bless v3.11.0
therefore it's the only version of Go that matters.
Reference issue: https://github.com/ooni/probe/issues/1738.
* fix(ptx/obfs4_test.go): avoid context-vs-normal-code race
We want to test whether we get the context failure if the error
generated inside normal code happens _after_ the context cancellation.
The best way to do that is to write code that is not racy. To this
end, we just need to pause normal code until we know that the context
has returned to the caller. We also need to ensure we do not leak
a goroutine, hence we use a WaitGroup to check that.
Fixes https://github.com/ooni/probe/issues/1750
When a probe gets a local DNS failure, it will continue and nonetheless
query the test helper without any IP address, just an empty list.
This diff fixes the behavior of cmd/oohelper to do the same.
Work part of https://github.com/ooni/probe/issues/1707.
Reference issue: https://github.com/ooni/probe/issues/1769
Motivation: The CI is failing. Those are integration tests. Let us figure out the issue when we approach release. Until we approach release, do not let those tests distracting us. Normal merges should only pass the `-short` tests.
This diff enables `websteps` to use uTLS for TLS parroting. It integrates the `oohttp.StdlibTransport` wrapper which uses the `ooni/oohttp` fork. `oohttp` supports TLS-like connections like `utls.Conn`.
As a prototype, the testhelper and `websteps` code now uses the `utls.HelloChrome_Auto` fingerprint, i.e. the simulated TLS fingerprint of the Google Chrome browser.
It is a further contribution for my GSoC project.
Reference issue: https://github.com/ooni/probe/issues/1733
This is the extension of https://github.com/ooni/probe-cli/pull/431, and my final deliverable for GSoC 2021.
The diff introduces:
1) The new `testhelper` which supports testing multiple IP endpoints per domain and introduces HTTP/3 control measurements. The specification of the `testhelper` can be found at https://github.com/ooni/spec/pull/219. The `testhelper` algorithm consists of three main steps:
* `InitialChecks` verifies that the input URL can be parsed, has an expected scheme, and contains a valid domain name.
* `Explore` enumerates all the URLs that it discovers by redirection from the original URL, or by detecting h3 support at the target host.
* `Generate` performs a step-by-step measurement of each discovered URL.
2) A prototype of the corresponding new experiment `websteps` which uses the control measurement of the `testhelper` to know which URLs to measure, and what to expect. The prototype does not yet have:
* unit and integration tests,
* an analysis tool to compare the control and the probe measurement.
This PR is my final deliverable as it is the outcome of the trials, considerations and efforts of my GSoC weeks at OONI.
It fully integrates HTTP/3 (QUIC) support which has been only used in the `urlgetter` experiment until now.
Related issues: https://github.com/ooni/probe/issues/1729 and https://github.com/ooni/probe/issues/1733.
The utility of SafeErrWrapperBuilder is low. Let us instead change the
code to always create ErrWrapper when we're in this package.
While there, also note some TODO-next items.
Part of https://github.com/ooni/probe/issues/1505.
* refactor(errorsx): start hiding private details and moving around stuff
Part of https://github.com/ooni/probe/issues/1505
* fix: remove now-addressed todo comments
* refactor: cleaner way of passing a UDPConn around
Also part of https://github.com/ooni/probe/issues/1505
* Update internal/engine/netx/quicdialer/connectionstate.go
I needed to add some tests as integration tests due to circular
imports, but this is ~fine because we quite likely want many
integration tests in the errorsx package anyway.
Part of https://github.com/ooni/probe/issues/1505.
With this factory, we want to construct ourselves the TLS dialer
so that we can use a dialer wrapper that always sets timeouts when
reading, addressing https://github.com/ooni/probe/issues/1609.
As a result, we cannot immediately replace the i/e/netx factory
for creating a new HTTP transport, since the functions signatures
are not directly compatible.
Refactoring is part of https://github.com/ooni/probe/issues/1505.
This diff is part of https://github.com/ooni/probe/issues/1505.
You will notice that I have not adapted all the (great) tests we had
previously. They should live at another layer, and namely the one that
deals with performing measurements.
When I'm refactoring such a layer I'll ensure those tests that I have
not adapted here are reintroduced into the tree.
Auto-configure every relevant TLS field as close as possible to
where it's actually used.
As a side effect, add support for mocking the creation of a TLS
connection, which should possibly be useful for uTLS?
Work that is part of https://github.com/ooni/probe/issues/1505
The BogonResolver relied on its wrapper resolver to pass along the
list of addresses _and_ the error. But the idiomatic thing to do is
often to return `nil` when there is an error.
I broke this very fragile assumption in https://github.com/ooni/probe-cli/pull/399.
I could of course fix it, but this assumption is clearly wrong
and we should not allow such fragile code in the tree.
We are not using BogonIsError much in the tree. The only place in
which we're using it for measuring seems to be dnscheck.
It may be that this surprising behavior was what caused the issue at
https://github.com/ooni/probe/issues/1510 in the first place.
Regardless, let's remove fragile code and adjust the test that was
failing. Also that test is quick so it can run in `-short` mode.
Spotted while working on https://github.com/ooni/probe/issues/1505.
* refactor(netxlite): make sure we always use netmocks
While there, improve logging and make sure we test 100% of the
package with unit tests only. (We don't need to have integration
testing in this package because it's fairly simple/obvious.)
Part of https://github.com/ooni/probe/issues/1505
* further improve logs
What do I mean by pivoting? Netx is currently organized by row:
```
| dialer | quicdialer | resolver | ...
saving | | | | ...
errorwrapping | | | | ...
logging | | | | ...
mocking/sys | | | | ...
```
Every row needs to implement saving, errorwrapping, logging, mocking (or
adapting to the system or to some underlying library).
This causes cross package dependencies and, in turn, complexity. For
example, we need the `trace` package for supporting saving.
And `dialer`, `quickdialer`, et al. need to depend on such a package.
The same goes for errorwrapping.
This arrangement further complicates testing. For example, I am
currently working on https://github.com/ooni/probe/issues/1505 and
I realize it need to repeat integration tests in multiple places.
Let's say instead we pivot the above matrix as follows:
```
| saving | errorwrapping | logging | ...
dialer | | | | ...
quicdialer | | | | ...
logging | | | | ...
mocking/sys | | | | ...
...
```
In this way, now every row contains everything related to a specific
action to perform. We can now share code without relying on extra
support packages. What's more, we can write tests and, judding from
the way in which things are made, it seems we only need integration
testing in `errorwrapping` because it's where data quality matters
whereas, in all other cases, unit testing is fine.
I am going, therefore, to proceed with these changes and "pivot"
`netx`. Hopefully, it won't be too painful.
We are not using them anymore. The only nettest still using the
legacy netx implementation is tor, for which setting these fields
is useless, because it performs each measurement into a separate
goroutine. Hence, let us start removing this part of the legacy
netx codebase, which is hampering progress in other areas.
Occurred to me while doing testing for the recent changes in
error mapping (https://github.com/ooni/probe/issues/1505).
* refactor: move scrubbingLogger to the scrubber pkg
We need it exported so we can use it in the new implementation.
Part of https://github.com/ooni/probe/issues/1687
* fix test
* refactor: move bytecounting conn in bytecounter pkg
This enables other pieces of code to request bytecounting without
depending on netx or on the perverse using-the-context-to-configure-
byte-counting mechanism.
Also occurred when working on https://github.com/ooni/probe/issues/1687
* fix: add missing docs
The current implementation assumes the user has already installed tor
on the current system. If tor is not present, the experiment fails.
This is meant to be the first version of this experiment.
We are going to add more functionality in subsequent revisions of
this experiment, once we've collected more feedback.
Reference issue: https://github.com/ooni/probe/issues/1565.
Here's the spec PR: https://github.com/ooni/spec/pull/218.
Here's the issue tracking future work: https://github.com/ooni/probe/issues/1686
This is a very light refactoring of the mlablocatev2 package where we do
the following things:
1. use interfaces rather than depending on other pkgs where possible
2. add a missing test to the test suite
3. write more comprehensive docs (including todo-next comments)
