Heavier network extensions might require the wireguard-go component to
use less ram, so let users of this reduce these as needed.
At some point we'll put this behind a configuration method of sorts, but
for now, just expose the consts as vars.
Requested-by: Josh Bleecher Snyder <[email protected]>
Signed-off-by: Jason A. Donenfeld <[email protected]>
On Linux we can run `ip link del wg0`, in which case the fd becomes
stale, and we should exit. Since this is an intentional action, don't
treat it as an error.
Signed-off-by: Jason A. Donenfeld <[email protected]>
The -1 protection was removed and the wrong error was returned, causing
us to read from a bogus fd. As well, remove the useless closures that
aren't doing anything, since this is all synchronized anyway.
Fixes: 10533c3 ("all: make conn.Bind.Open return a slice of receive functions")
Signed-off-by: Jason A. Donenfeld <[email protected]>
When retrying, if count is not 0, we forget to dequeue another request,
and so the ring fills up and errors out.
Reported-by: Sascha Dierberg <[email protected]>
Signed-off-by: Jason A. Donenfeld <[email protected]>
In 097af6e ("tun: windows: protect reads from closing") we made sure no
functions are running when End() is called, to avoid a UaF. But we still
need to kick that event somehow, so that Read() is allowed to exit, in
order to release the lock. So this commit calls SetEvent, while moving
the closing boolean to be atomic so it can be modified without locks,
and then moves to a WaitGroup for the RCU-like pattern.
Signed-off-by: Jason A. Donenfeld <[email protected]>
The code previously used the old errors channel for checking, rather
than the simpler boolean, which caused issues on shutdown, since the
errors channel was meaningless. However, looking at this exposed a more
basic problem: Close() and all the other functions that check the closed
boolean can race. So protect with a basic RW lock, to ensure that
Close() waits for all pending operations to complete.
Reported-by: Joshua Sjoding <[email protected]>
Signed-off-by: Jason A. Donenfeld <[email protected]>
If we receive a large UDP packet, don't return an error to receive.go,
which then terminates the receive loop. Instead, simply retry.
Considering Winsock's general finickiness, we might consider other
places where an attacker on the wire can generate error conditions like
this.
Reported-by: Sascha Dierberg <[email protected]>
Signed-off-by: Jason A. Donenfeld <[email protected]>
When we pass the TUN FD to the child, we have to call TUNSIFPID;
otherwise when we close the device, we get a splat in dmesg.
Signed-off-by: Jason A. Donenfeld <[email protected]>
By not comparing these with the modulo, the ring became nearly never
full, resulting in completion queue buffers filling up prematurely.
Reported-by: Joshua Sjoding <[email protected]>
Signed-off-by: Jason A. Donenfeld <[email protected]>
There's no way for len(peers)==0 when a current peer has
isRunning==false.
This requires some struct reshuffling so that the uint64 pointer is
aligned.
Signed-off-by: Jason A. Donenfeld <[email protected]>
Googlers have a habit of graffiting their name in TODO items that then
are never addressed, and other people won't go near those because
they're marked territory of another animal. I've been gradually cleaning
these up as I see them, but this commit just goes all the way and
removes the remaining stragglers.
Signed-off-by: Jason A. Donenfeld <[email protected]>
This code is stable, and the test is finicky, especially on high core
count systems, so just disable it.
Signed-off-by: Jason A. Donenfeld <[email protected]>
This linked list implementation is awful, but maybe Go 2 will help
eventually, and at least we're not open coding the hlist any more.
Signed-off-by: Jason A. Donenfeld <[email protected]>
We're loosing our ownership of the port when bringing the device down,
which means another test process could reclaim it. Avoid this by
retrying for 4 seconds.
Signed-off-by: Jason A. Donenfeld <[email protected]>
Rather than racing with Start(), since we're never destroying these
queues, we just set the variables at creation time.
Signed-off-by: Jason A. Donenfeld <[email protected]>
Since RoutineHandshake calls peer.SendKeepalive(), it potentially is a
writer into the encryption queue, so we need to bump the wg count.
Signed-off-by: Jason A. Donenfeld <[email protected]>
It's never used and we won't have a use for it. Also, move to go-running
stringer, for those without GOPATHs.
Signed-off-by: Jason A. Donenfeld <[email protected]>
The test previously checked the offset within a substruct, not the
offset within the allocated struct, so this adds the two together.
It then fixes an alignment crash on 32-bit machines.
Signed-off-by: Jason A. Donenfeld <[email protected]>
Part of being actually idempotent is that we shouldn't penalize code
that takes advantage of this property with a log splat.
Signed-off-by: Jason A. Donenfeld <[email protected]>
Before, the code attached a finalizer to an object that wasn't returned,
resulting in immediate garbage collection. Instead return the actual
pointer.
Signed-off-by: Jason A. Donenfeld <[email protected]>
Without this, we wind up freeing packets that the encryption/decryption
queues still have, resulting in a UaF.
Signed-off-by: Jason A. Donenfeld <[email protected]>
Here is the old implementation:
type WaitPool struct {
c chan interface{}
}
func NewWaitPool(max uint32, new func() interface{}) *WaitPool {
p := &WaitPool{c: make(chan interface{}, max)}
for i := uint32(0); i < max; i++ {
p.c <- new()
}
return p
}
func (p *WaitPool) Get() interface{} {
return <- p.c
}
func (p *WaitPool) Put(x interface{}) {
p.c <- x
}
It performs worse than the new one:
name old time/op new time/op delta
WaitPool-16 16.4µs ± 5% 15.1µs ± 3% -7.86% (p=0.008 n=5+5)
Signed-off-by: Jason A. Donenfeld <[email protected]>
benchmark old ns/op new ns/op delta
BenchmarkUAPIGet-16 2872 2157 -24.90%
benchmark old allocs new allocs delta
BenchmarkUAPIGet-16 30 18 -40.00%
benchmark old bytes new bytes delta
BenchmarkUAPIGet-16 737 256 -65.26%
Signed-off-by: Jason A. Donenfeld <[email protected]>
This moves to a simple queue with no routine processing it, to reduce
scheduler pressure.
This splits latency in half!
benchmark old ns/op new ns/op delta
BenchmarkThroughput-16 2394 2364 -1.25%
BenchmarkLatency-16 259652 120810 -53.47%
Signed-off-by: Jason A. Donenfeld <[email protected]>
This makes the IpcGet method much faster.
We also refactor the traversal API to use a callback so that we don't
need to allocate at all. Avoiding allocations we do self-masking on
insertion, which in turn means that split intermediate nodes require a
copy of the bits.
benchmark old ns/op new ns/op delta
BenchmarkUAPIGet-16 3243 2659 -18.01%
benchmark old allocs new allocs delta
BenchmarkUAPIGet-16 35 30 -14.29%
benchmark old bytes new bytes delta
BenchmarkUAPIGet-16 1218 737 -39.49%
This benchmark is good, though it's only for a pair of peers, each with
only one allowedips. As this grows, the delta expands considerably.
Signed-off-by: Jason A. Donenfeld <[email protected]>
Jason A. Donenfeld