Without any synchronization here, the "first packet" callback can
be delayed indefinitely, while other work continues.
Since the callback starts the benchmark timer, this could skew results.
Worse, if the benchmark manages to complete before the benchmark timer begins,
it'll cause a data race with the benchmark shutdown performed by package testing.
That is what is reported in #1881.
This is a bit unfortunate, in that it means that users of TrafficGen have
to be careful to keep this callback speedy and lightweight and to avoid deadlocks.
Fixes#1881
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
It is possible to get multiple status callbacks from an Engine.
We need to wait for at least one from each Engine.
Without limiting to one per Engine,
wait.Wait can exit early or can panic due to a negative counter.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
This reduces the speed with which these benchmarks exhaust their supply fds.
Not to zero unfortunately, but it's still helpful when doing long runs.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
This function accounted for ~1% of all allocs by tailscaled.
It is trivial to improve, so may as well.
name old time/op new time/op delta
KeyMarshalText-8 197ns ± 0% 47ns ± 0% -76.12% (p=0.016 n=4+5)
name old alloc/op new alloc/op delta
KeyMarshalText-8 200B ± 0% 80B ± 0% -60.00% (p=0.008 n=5+5)
name old allocs/op new allocs/op delta
KeyMarshalText-8 5.00 ± 0% 1.00 ± 0% -80.00% (p=0.008 n=5+5)
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
The old way was way too fragile and had felt like it had more special
cases than normal cases. (see #1874, #1860, #1834, etc) It became very
obvious the old algorithm didn't work when we made the output be
pretty and try to show the user the command they need to run in
5ecc7c7200 for #1746)
The new algorithm is to map the prefs (current and new) back to flags
and then compare flags. This nicely handles the OS-specific flags and
the n:1 and 1:n flag:pref cases.
No change in the existing already-massive test suite, except some ordering
differences (the missing items are now sorted), but some new tests are
added for behavior that was broken before. In particular, it now:
* preserves non-pref boolean flags set to false, and preserves exit
node IPs (mapping them back from the ExitNodeID pref, as well as
ExitNodeIP),
* doesn't ignore --advertise-exit-node when doing an EditPrefs call
(#1880)
* doesn't lose the --operator on the non-EditPrefs paths (e.g. with
--force-reauth, or when the backend was not in state Running).
Fixes#1880
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Needed for the "up checker" to map back from exit node stable IDs (the
ipn.Prefs.ExitNodeID) back to an IP address in error messages.
But also previously requested so people can use it to then make API
calls. The upcoming "tailscale admin" subcommand will probably need it
too.
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
This reverts commit 7d16c8228b.
I have no idea how I ended up here. The bug I was fixing with this change
fails to reproduce on Ubuntu 18.04 now, and this change definitely does
break 20.04, 20.10, and Debian Buster. So, until we can reliably reproduce
the problem this was meant to fix, reverting.
Part of #1875
Signed-off-by: David Anderson <dave@natulte.net>
The --advertise-routes and --advertise-exit-node flags both mutating
one pref is the gift that keeps on giving.
I need to rewrite the this up warning code to first map prefs back to
flag values and then just compare flags instead of comparing prefs,
but this is the minimal fix for now.
This also includes work on the tests, to make them easier to write
(and more accurate), by letting you write the flag args directly and
have that parse into the upArgs/MaskedPrefs directly, the same as the
code, rather than them being possibly out of sync being written by
hand.
Fixes https://twitter.com/EXPbits/status/1390418145047887877
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Fields rename only.
Part of the general effort to make our code agnostic about endpoint formatting.
It's just a name, but it will soon be a misleading one; be more generic.
Do this as a separate commit because it generates a lot of whitespace changes.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
Upstream wireguard-go renamed the interface method
from CreateEndpoint to ParseEndpoint.
I missed some comments. Fix them.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
By using conn.NewDefaultBind, this test requires that our endpoints
be comprehensible to wireguard-go. Instead, use a no-op bind that
treats endpoints as opaque strings.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
Instead of calling ParseHex, do the hex.Decode directly.
name old time/op new time/op delta
UnmarshalJSON-8 86.9ns ± 0% 42.6ns ± 0% -50.94% (p=0.000 n=15+14)
name old alloc/op new alloc/op delta
UnmarshalJSON-8 128B ± 0% 0B -100.00% (p=0.000 n=15+15)
name old allocs/op new allocs/op delta
UnmarshalJSON-8 2.00 ± 0% 0.00 -100.00% (p=0.000 n=15+15)
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
Legacy endpoints (addrSet) currently reconstruct their dst string when requested.
Instead, store the dst string we were given to begin with.
In addition to being simpler and cheaper, this makes less code
aware of how to interpret endpoint strings.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
Prefer the error from the actual wireguard-go device method call,
not {To,From}UAPI, as those tend to be less interesting I/O errors.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
When wireguard-go's UAPI interface fails with an error, ReconfigDevice hangs.
Fix that by buffering the channel and closing the writer after the call.
The code now matches the corresponding code in DeviceConfig, where I got it right.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
It is unused, and has been since early Feb 2021 (Tailscale 1.6).
We can't get delete the DeviceOptions entirely yet;
first #1831 and #1839 need to go in, along with some wireguard-go changes.
Deleting this chunk of code now will make the later commits more clearly correct.
Pingers can now go too.
Signed-off-by: Josh Bleecher Snyder <josh@tailscale.com>
The earlier eb06ec172f fixed
the flaky SSH issue (tailscale/corp#1725) by making sure that packets
addressed to Tailscale IPs in hybrid netstack mode weren't delivered
to netstack, but another issue remained:
All traffic handled by netstack was also potentially being handled by
the host networking stack, as the filter hook returned "Accept", which
made it keep processing. This could lead to various random racey chaos
as a function of OS/firewalls/routes/etc.
Instead, once we inject into netstack, stop our caller's packet
processing.
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Whenever we dropped a packet due to ACLs, wireguard-go was logging:
Failed to write packet to TUN device: packet dropped by filter
Instead, just lie to wireguard-go and pretend everything is okay.
Fixes#1229
Signed-off-by: Brad Fitzpatrick <bradfitz@tailscale.com>
Upstream wireguard-go renamed the interface method
from CreateEndpoint to ParseEndpoint.
I updated the log call site but not the allowlist.
Signed-off-by: Josh Bleecher Snyder <josharian@gmail.com>
#1817 removed the only place in our CI where we executed our benchmark code.
Fix that by executing it everywhere.
The benchmarks are generally cheap and fast,
so this should add minimal overhead.
Signed-off-by: Josh Bleecher Snyder <josharian@gmail.com>
To prevent issues like #1786, run staticcheck on the primary GOOSes:
linux, mac, and windows.
Windows also has a fair amount of GOARCH-specific code.
If we ever have GOARCH staticcheck failures on other GOOSes,
we can expand the test matrix further.
This requires installing the staticcheck binary so that
we can execute it with different GOOSes.
Signed-off-by: Josh Bleecher Snyder <josharian@gmail.com>
This is needed because the original opts.Prefs field was at some point
subverted for use in frontend->backend state migration for backward
compatibility on some platforms. We still need that feature, but we
also need the feature of providing the full set of prefs from
`tailscale up`, *not* including overwriting the prefs.Persist keys, so
we can't use the original field from `tailscale up`.
`tailscale up` had attempted to compensate for that by doing SetPrefs()
before Start(), but that violates the ipn.Backend contract, which says
you should call Start() before anything else (that's why it's called
Start()). As a result, doing SetPrefs({ControlURL=...,
WantRunning=true}) would cause a connection to the *previous* control
server (because WantRunning=true), and then connect to the *new*
control server only after running Start().
This problem may have been avoided before, but only by pure luck.
It turned out to be relatively harmless since the connection to the old
control server was immediately closed and replaced anyway, but it
created a race condition that could have caused spurious notifications
or rejected keys if the server responded quickly.
As already covered by existing TODOs, a better fix would be to have
Start() get out of the business of state migration altogether. But
we're approaching a release so I want to make the minimum possible fix.
Fixes#1840.
Signed-off-by: Avery Pennarun <apenwarr@tailscale.com>
We were over-eager in running tailscale in GUI mode.
f42ded7acf fixed that by
checking for a variety of shell-ish env vars and using those
to force us into CLI mode.
However, for reasons I don't understand, those shell env vars
are present when Xcode runs Tailscale.app on my machine.
(I've changed no configs, modified nothing on a brand new machine.)
Work around that by adding an additional "only in GUI mode" check.
Signed-off-by: Josh Bleecher Snyder <josharian@gmail.com>
