David Crawshaw [Tue, 7 Apr 2020 04:52:17 +0000 (14:52 +1000)]
device: move stats fields back down and add test diagnostics
This reverts the movement of fields from d49f4e9.
That commit was cherry-picked from another branch where a field
had changed and misaligned the atomic fields. After cherry-picking,
moving the fields was no longer necessary but got dragged along.
Signed-off-by: David Crawshaw <crawshaw@tailscale.com>
David Anderson [Wed, 11 Mar 2020 00:52:50 +0000 (17:52 -0700)]
device: remove racey read in session key rotation.
This code was attempting to use the "compare racily, then lock
and compare again" idiom to try and reduce lock contention.
However, that idiom is not safe to use unless the comparison
uses atomic operations, which this does not.
This change simply deletes the racy read. This makes the code
correct, but potentially increases lock contention.
Signed-off-by: David Anderson <danderson@tailscale.com>
Avery Pennarun [Wed, 6 Nov 2019 08:28:02 +0000 (00:28 -0800)]
tun: NetlinkListener: don't send EventDown before sending EventUp
This works around a startup race condition when competing with
HackListener, which is trying to do the same job. If HackListener
detects that the tundev is running while there is still an event in the
netlink queue that says it isn't running, then the device receives a
string of events like
EventUp (HackListener)
EventDown (NetlinkListener)
EventUp (NetlinkListener)
Unfortunately, after the first EventDown, the device stops itself,
thinking incorrectly that the administrator has downed its tundev.
The device is ignoring the initial EventDown anyway, so just don't emit
it.
David Anderson [Sun, 1 Mar 2020 08:39:24 +0000 (00:39 -0800)]
device: make Peer fields safe for atomic access on 32-bit.
All atomic access must be aligned to 64 bits, even on 32-bit
platforms. Go promises that the start of allocated structs is
aligned to 64 bits. So, place the atomically-accessed things
first in the struct so that they benefit from that alignment.
As a side bonus, it cleanly separates fields that are accessed
by atomic ops, and those that should be accessed under mu.
Also adds a test that will fail consistently on 32-bit platforms
if the struct ever changes again to violate the rules. This is
likely not needed because unaligned access crashes reliably,
but this will reliably fail even if tests accidentally pass due
to lucky alignment.
Signed-Off-By: David Anderson <danderson@tailscale.com>
David Crawshaw [Wed, 19 Feb 2020 15:09:24 +0000 (10:09 -0500)]
rwcancel: no-op builds for windows and darwin
This lets us include the package on those platforms in a
followup commit where we split out a conn package from device.
It also lets us run `go test ./...` when developing on macOS.
Signed-off-by: David Crawshaw <crawshaw@tailscale.com>
David Crawshaw [Sun, 8 Dec 2019 23:22:31 +0000 (18:22 -0500)]
ratelimiter: use a fake clock in tests and style cleanups
The existing test would occasionally flake out with:
--- FAIL: TestRatelimiter (0.12s)
ratelimiter_test.go:99: Test failed for 127.0.0.1 , on: 7 ( not having refilled enough ) expected: false got: true
FAIL
FAIL golang.zx2c4.com/wireguard/ratelimiter 0.171s
The fake clock also means the tests run much faster, so
testing this package with -count=1000 now takes < 100ms.
While here, several style cleanups. The most significant one
is unembeding the sync.Mutex fields in the rate limiter objects.
Embedded as they were, the lock methods were accessible
outside the ratelimiter package. As they aren't needed externally,
keep them internal to make them easier to reason about.
Passes `go test -race -count=10000 ./ratelimiter`
Signed-off-by: David Crawshaw <crawshaw@tailscale.com>
Tobias Klauser [Wed, 4 Mar 2020 16:21:54 +0000 (17:21 +0100)]
global: use RTMGRP_* consts from x/sys/unix
Update the golang.org/x/sys/unix dependency and use the newly introduced
RTMGRP_* consts instead of using the corresponding RTNLGRP_* const to
create a mask.
Some devices take ~2 seconds to enumerate on Windows if we try to get
their instance name. The hardware id property, on the other hand,
is available right away.
Signed-off-by: Avery Pennarun <apenwarr@gmail.com>
[zx2c4: inlined this to where it makes sense, reused setupapi const]
Avery Pennarun [Sat, 12 Oct 2019 07:46:13 +0000 (00:46 -0700)]
rwcancel: handle EINTR and EAGAIN in unixSelect()
On my Chromebook (Linux 4.19.44 in a VM) and on an AWS EC2
machine, select() was sometimes returning EINTR. This is
harmless and just means you should try again. So let's try
again.
This eliminates a problem where the tunnel fails to come up
correctly and the program needs to be restarted.
tun: openbsd: check for interface already being up
In some cases, we operate on an already-up interface, or the user brings
up the interface before we start monitoring. For those situations, we
should first check if the interface is already up.
This still technically races between the initial check and the start of
the route loop, but fixing that is a bit ugly and probably not worth it
at the moment.
Simon Rozman [Wed, 28 Aug 2019 09:39:01 +0000 (11:39 +0200)]
wintun: upgrade deleting all interfaces and make it reusable
DeleteAllInterfaces() didn't check if SPDRP_DEVICEDESC == "WireGuard
Tunnel". It deleted _all_ Wintun adapters, not just WireGuard's.
Furthermore, the DeleteAllInterfaces() was upgraded into a new function
called DeleteMatchingInterfaces() for selectively deletion. This will
be used by WireGuard to clean stale Wintun adapters.
projects / thirdparty / wireguard-go.git / log
