This change supports using a user supplied TCP MSS for new active TCP
connections. Note, the user supplied MSS must be less than or equal to the
maximum possible MSS for a TCP connection's route. If it is greater than the
maximum possible MSS, the maximum possible MSS will be used as the connection's
MSS instead.
This change does not use this user supplied MSS for connections accepted from
listening sockets - that will come in a later change.
Test: Test that outgoing TCP SYN segments contain a TCP MSS option with the user
supplied MSS if it is not greater than the maximum possible MSS for the route.
PiperOrigin-RevId: 277185125
This change makes it so that NDP work is done using the per-interface NDP
configurations instead of the stack-wide default NDP configurations to correctly
implement RFC 4861 section 6.3.2 (note here, a host is a single NIC operating
as a host device), and RFC 4862 section 5.1.
Test: Test that we can set NDP configurations on a per-interface basis without
affecting the configurations of other interfaces or the stack-wide default. Also
make sure that after the configurations are updated, the updated configurations
are used for NDP processes (e.g. Duplicate Address Detection).
PiperOrigin-RevId: 276525661
This change introduces a new interface, stack.NDPDispatcher. It can be
implemented by the netstack integrator to receive NDP related events. As of this
change, only DAD related events are supported.
Tests: Existing tests were modified to use the NDPDispatcher's DAD events for
DAD tests where it needed to wait for DAD completing (failing and resolving).
PiperOrigin-RevId: 276338733
This change makes sure that when an address which is already known by a NIC and
has kind = permanentExpired gets promoted to permanent, the new
PrimaryEndpointBehavior is respected.
PiperOrigin-RevId: 276136317
Right now, we send each tcp packet separately, we call one system
call per-packet. This patch allows to generate multiple tcp packets
and send them by sendmmsg.
The arguable part of this CL is a way how to handle multiple headers.
This CL adds the next field to the Prepandable buffer.
Nginx test results:
Server Software: nginx/1.15.9
Server Hostname: 10.138.0.2
Server Port: 8080
Document Path: /10m.txt
Document Length: 10485760 bytes
w/o gso:
Concurrency Level: 5
Time taken for tests: 5.491 seconds
Complete requests: 100
Failed requests: 0
Total transferred: 1048600200 bytes
HTML transferred: 1048576000 bytes
Requests per second: 18.21 [#/sec] (mean)
Time per request: 274.525 [ms] (mean)
Time per request: 54.905 [ms] (mean, across all concurrent requests)
Transfer rate: 186508.03 [Kbytes/sec] received
sw-gso:
Concurrency Level: 5
Time taken for tests: 3.852 seconds
Complete requests: 100
Failed requests: 0
Total transferred: 1048600200 bytes
HTML transferred: 1048576000 bytes
Requests per second: 25.96 [#/sec] (mean)
Time per request: 192.576 [ms] (mean)
Time per request: 38.515 [ms] (mean, across all concurrent requests)
Transfer rate: 265874.92 [Kbytes/sec] received
w/o gso:
$ ./tcp_benchmark --client --duration 15 --ideal
[SUM] 0.0-15.1 sec 2.20 GBytes 1.25 Gbits/sec
software gso:
$ tcp_benchmark --client --duration 15 --ideal --gso $((1<<16)) --swgso
[SUM] 0.0-15.1 sec 3.99 GBytes 2.26 Gbits/sec
PiperOrigin-RevId: 276112677
This change adds support for optionally auto-generating an IPv6 link-local
address based on the NIC's MAC Address on NIC enable.
Note, this change will not break existing uses of netstack as the default
configuration for the stack options is set in such a way that a link-local
address will not be auto-generated unless the stack is explicitly configured.
See `stack.Options` for more details. Specifically, see
`stack.Options.AutoGenIPv6LinkLocal`.
Tests: Tests to make sure that the IPb6 link-local address is only
auto-generated if the stack is specifically configured to do so. Also tests to
make sure that an auto-generated address goes through the DAD process.
PiperOrigin-RevId: 276059813
Like (AF_INET, SOCK_RAW) sockets, AF_PACKET sockets require CAP_NET_RAW. With
runsc, you'll need to pass `--net-raw=true` to enable them.
Binding isn't supported yet.
PiperOrigin-RevId: 275909366
It is quite legal to send from the ANY address (it is required for
DHCP). I can't figure out why the broadcast address was included here,
so removing that as well.
PiperOrigin-RevId: 275541954
NDP Neighbor Solicitations sent during Duplicate Address Detection must have an
IP hop limit of 255, as all NDP Neighbor Solicitations should have.
Test: Test that DAD messages have the IPv6 hop limit field set to 255.
PiperOrigin-RevId: 275321680
This change adds support for Duplicate Address Detection on IPv6 addresses
as defined by RFC 4862 section 5.4.
Note, this change will not break existing uses of netstack as the default
configuration for the stack options is set in such a way that DAD will not be
performed. See `stack.Options` and `stack.NDPConfigurations` for more details.
Tests: Tests to make sure that the DAD process properly resolves or fails.
That is, tests make sure that DAD resolves only if:
- No other node is performing DAD for the same address
- No other node owns the same address
PiperOrigin-RevId: 275189471
...and do not populate link address cache at dispatch. This partially
reverts 313c767b00, which caused malformed
packets (e.g. NDP Neighbor Adverts with incorrect hop limit values) to
populate the address cache. In particular, this masked a bug that was
introduced to the Neighbor Advert generation code in
7c1587e340.
PiperOrigin-RevId: 274865182
The behavior for sending and receiving local broadcast (255.255.255.255)
traffic is as follows:
Outgoing
--------
* A broadcast packet sent on a socket that is bound to an interface goes out
that interface
* A broadcast packet sent on an unbound socket follows the route table to
select the outgoing interface
+ if an explicit route entry exists for 255.255.255.255/32, use that one
+ else use the default route
* Broadcast packets are looped back and delivered following the rules for
incoming packets (see next). This is the same behavior as for multicast
packets, except that it cannot be disabled via sockopt.
Incoming
--------
* Sockets wishing to receive broadcast packets must bind to either INADDR_ANY
(0.0.0.0) or INADDR_BROADCAST (255.255.255.255). No other socket receives
broadcast packets.
* Broadcast packets are multiplexed to all sockets matching it. This is the
same behavior as for multicast packets.
* A socket can bind to 255.255.255.255:<port> and then receive its own
broadcast packets sent to 255.255.255.255:<port>
In addition, this change implicitly fixes an issue with multicast reception. If
two sockets want to receive a given multicast stream and one is bound to ANY
while the other is bound to the multicast address, only one of them will
receive the traffic.
PiperOrigin-RevId: 272792377
Netstack always picks a random start point everytime PickEphemeralPort
is called. While this is required for UDP so that DNS requests go
out through a randomized set of ports it is not required for TCP. Infact
Linux explicitly hashes the (srcip, dstip, dstport) and a one time secret
initialized at start of the application to get a random offset. But to
ensure it doesn't start from the same point on every scan it uses a static
hint that is incremented by 2 in every call to pick ephemeral ports.
The reason for 2 is Linux seems to split the port ranges where active connects
seem to use even ones while odd ones are used by listening sockets.
This CL implements a similar strategy where we use a hash + hint to generate
the offset to start the search for a free Ephemeral port.
This ensures that we cycle through the available port space in order for
repeated connects to the same destination and significantly reduces the
chance of picking a recently released port.
PiperOrigin-RevId: 272058370
Non-primary addresses are used for endpoints created to accept multicast and
broadcast packets, as well as "helper" endpoints (0.0.0.0) that allow sending
packets when no proper address has been assigned yet (e.g., for DHCP). These
addresses are not real addresses from a user point of view and should not be
part of the NICInfo() value. Also see b/127321246 for more info.
This switches NICInfo() to call a new NIC.PrimaryAddresses() function. To still
allow an option to get all addresses (mostly for testing) I added
Stack.GetAllAddresses() and NIC.AllAddresses().
In addition, the return value for GetMainNICAddress() was changed for the case
where the NIC has no primary address. Instead of returning an error here,
it now returns an empty AddressWithPrefix() value. The rational for this
change is that it is a valid case for a NIC to have no primary addresses.
Lastly, I refactored the code based on the new additions.
PiperOrigin-RevId: 270971764
Previously, the only safe way to use an fdbased endpoint was to leak the FD.
This change makes it possible to safely close the FD.
This is the first step towards having stoppable stacks.
Updates #837
PiperOrigin-RevId: 270346582
This also allows the tee(2) implementation to be enabled, since dup can now be
properly supported via WriteTo.
Note that this change necessitated some minor restructoring with the
fs.FileOperations splice methods. If the *fs.File is passed through directly,
then only public API methods are accessible, which will deadlock immediately
since the locking is already done by fs.Splice. Instead, we pass through an
abstract io.Reader or io.Writer, which elide locks and use the underlying
fs.FileOperations directly.
PiperOrigin-RevId: 268805207
The IPv6 all-nodes multicast address will be joined on NIC enable, and the
appropriate IPv6 solicited-node multicast address will be joined when IPv6
addresses are added.
Tests: Test receiving packets destined to the IPv6 link-local all-nodes
multicast address and the IPv6 solicted node address of an added IPv6 address.
PiperOrigin-RevId: 268047073
This also renames "subnet" to "addressRange" to avoid any more confusion with
an interface IP's subnet.
Lastly, this also removes the Stack.ContainsSubnet(..) API since it isn't used
by anyone. Plus the same information can be obtained from
Stack.NICAddressRanges().
PiperOrigin-RevId: 267229843
Adds support to generate Port Unreachable messages for UDP
datagrams received on a port for which there is no valid
endpoint.
Fixes#703
PiperOrigin-RevId: 267034418
This allows the stack to learn remote link addresses on incoming
packets, reducing the need to ARP to send responses.
This also reduces the number of round trips to the system clock,
since that may also prove to be performance-sensitive.
Fixes#739.
PiperOrigin-RevId: 265815816
This addresses the problem where an endpoint has its address removed but still
has outstanding references held by routes used in connected TCP/UDP sockets
which prevent the removal of the endpoint.
The fix adds a new "expired" flag to the referenced network endpoint, which is
set when an endpoint has its address removed. Incoming packets are not
delivered to an expired endpoint (unless in promiscuous mode), while sending
outgoing packets triggers an error to the caller (unless in spoofing mode).
In addition, a few helper functions were added to stack_test.go to reduce
code duplications.
PiperOrigin-RevId: 265514326
This adds the same logic to NIC.findEndpoint that is already done in
NIC.getRef. Since this makes the two functions very similar they were combined
into one with the originals being wrappers.
PiperOrigin-RevId: 263864708
Endpoint protocol goroutines were previously started as part of
loading the endpoint. This is potentially too soon, as resources used
by these goroutine may not have been loaded. Protocol goroutines may
perform meaningful work as soon as they're started (ex: incoming
connect) which can cause them to indirectly access resources that
haven't been loaded yet.
This CL defers resuming all protocol goroutines until the end of
restore.
PiperOrigin-RevId: 262409429
This allows the user code to add a network address with a subnet prefix length.
The prefix length value is stored in the network endpoint and provided back to
the user in the ProtocolAddress type.
PiperOrigin-RevId: 259807693
iptables also relies on IPPROTO_RAW in a way. It opens such a socket to
manipulate the kernel's tables, but it doesn't actually use any of the
functionality. Blegh.
PiperOrigin-RevId: 257903078
The implementation is similar to linux where we track the number of bytes
consumed by the application to grow the receive buffer of a given TCP endpoint.
This ensures that the advertised window grows at a reasonable rate to accomodate
for the sender's rate and prevents large amounts of data being held in stack
buffers if the application is not actively reading or not reading fast enough.
The original paper that was used to implement the linux receive buffer auto-
tuning is available @ https://public.lanl.gov/radiant/pubs/drs/lacsi2001.pdf
NOTE: Linux does not implement DRS as defined in that paper, it's just a good
reference to understand the solution space.
Updates #230
PiperOrigin-RevId: 253168283
Ghanan Gowripalan
Andrei Vagin
Kevin Krakauer
Tamir Duberstein
gVisor bot
Ian Gudger
Chris Kuiper
Bhasker Hariharan
Adin Scannell
Michael Pratt
Rahat Mahmood