This code path is for outgoing packets, and we don't currently do memory
accounting on this path. So it wasn't breaking anything.
This change did not add a test for ref-counting issue fixed, but will switch to
the leak-checking ref-counter later when all ref-counting issues are fixed.
PiperOrigin-RevId: 373447913
This is done for IPv4, UDP and TCP headers.
This also changes the packet checkers used in tests to error on
zero-checksum, not sure why it was allowed before.
And while I'm here, make comments' case consistent.
RELNOTES: n/a
Fixes#5049
PiperOrigin-RevId: 369383862
- Implement Stringer for it so that we can improve error messages.
- Use TCPFlags through the code base. There used to be a mixed usage of byte,
uint8 and int as TCP flags.
PiperOrigin-RevId: 361940150
One of the preparation to decouple underlying buffer implementation.
There are still some methods that tie to VectorisedView, and they will be
changed gradually in later CLs.
This CL also introduce a new ICMPv6ChecksumParams to replace long list of
parameters when calling ICMPv6Checksum, aiming to be more descriptive.
PiperOrigin-RevId: 360778149
Link address resolution is performed at the link layer (if required) so
we can defer it from the transport layer. When link resolution is
required, packets will be queued and sent once link resolution
completes. If link resolution fails, the transport layer will receive a
control message indicating that the stack failed to route the packet.
tcpip.Endpoint.Write no longer returns a channel now that writes do not
wait for link resolution at the transport layer.
tcpip.ErrNoLinkAddress is no longer used so it is removed.
Removed calls to stack.Route.ResolveWith from the transport layer so
that link resolution is performed when a route is created in response
to an incoming packet (e.g. to complete TCP handshakes or send a RST).
Tests:
- integration_test.TestForwarding
- integration_test.TestTCPLinkResolutionFailure
Fixes#4458
RELNOTES: n/a
PiperOrigin-RevId: 351684158
Read now takes a destination io.Writer, count, options. Keeping the method name
Read, in contrast to the Write method.
This enables:
* direct transfer of views under VV
* zero copy
It also eliminates the need for sentry to keep a slice of view because
userspace had requested a read that is smaller than the view returned, removing
the complexity there.
Read/Peek/ReadPacket are now consolidated together and some duplicate code is
removed.
PiperOrigin-RevId: 350636322
With the recent changes db36d948fa, we try
to balance the receive window advertisements between payload lengths vs
segment overhead length. This works fine when segment size are much
higher than the overhead, but not otherwise. In cases where the segment
length is smaller than the segment overhead, we may end up not
advertising zero receive window for long time and end up tail-dropping
segments. This is especially pronounced when application socket reads
are slow or stopped. In this change we do not grow the right edge of
the receive window for smaller segment sizes similar to Linux.
Also, we keep track of the socket buffer usage and let the window grow
if the application is actively reading data.
Fixes#4903
PiperOrigin-RevId: 345832012
* Remove stack.Route from incoming packet path.
There is no need to pass around a stack.Route during the incoming path
of a packet. Instead, pass around the packet's link/network layer
information in the packet buffer since all layers may need this
information.
* Support address bound and outgoing packet NIC in routes.
When forwarding is enabled, the source address of a packet may be bound
to a different interface than the outgoing interface. This change
updates stack.Route to hold both NICs so that one can be used to write
packets while the other is used to check if the route's bound address
is valid. Note, we need to hold the address's interface so we can check
if the address is a spoofed address.
* Introduce the concept of a local route.
Local routes are routes where the packet never needs to leave the stack;
the destination is stack-local. We can now route between interfaces
within a stack if the packet never needs to leave the stack, even when
forwarding is disabled.
* Always obtain a route from the stack before sending a packet.
If a packet needs to be sent in response to an incoming packet, a route
must be obtained from the stack to ensure the stack is configured to
send packets to the packet's source from the packet's destination.
* Enable spoofing if a stack may send packets from unowned addresses.
This change required changes to some netgophers since previously,
promiscuous mode was enough to let the netstack respond to all
incoming packets regardless of the packet's destination address. Now
that a stack.Route is not held for each incoming packet, finding a route
may fail with local addresses we don't own but accepted packets for
while in promiscuous mode. Since we also want to be able to send from
any address (in response the received promiscuous mode packets), we need
to enable spoofing.
* Skip transport layer checksum checks for locally generated packets.
If a packet is locally generated, the stack can safely assume that no
errors were introduced while being locally routed since the packet is
never sent out the wire.
Some bugs fixed:
- transport layer checksum was never calculated after NAT.
- handleLocal didn't handle routing across interfaces.
- stack didn't support forwarding across interfaces.
- always consult the routing table before creating an endpoint.
Updates #4688Fixes#3906
PiperOrigin-RevId: 340943442
RACK detects packet reordering by checking if the sender received ACK for
the packet which has the sequence number less than the already acknowledged
packets.
PiperOrigin-RevId: 336397526
segment_queue today has its own standalone limit of MaxUnprocessedSegments but
this can be a problem in UnlockUser() we do not release the lock till there are
segments to be processed. What can happen is as handleSegments dequeues packets
more keep getting queued and we will never release the lock. This can keep
happening even if the receive buffer is full because nothing can read() till we
release the lock.
Further having a separate limit for pending segments makes it harder to track
memory usage etc. Unifying the limits makes it easier to reason about memory in
use and makes the overall buffer behaviour more consistent.
PiperOrigin-RevId: 333508122
RACK requires the segments to be in the order of their transmission
or retransmission times. This cl creates a new list and moves the
retransmitted segments to the end of the list.
PiperOrigin-RevId: 327325153
Formerly, when a packet is constructed or parsed, all headers are set by the
client code. This almost always involved prepending to pk.Header buffer or
trimming pk.Data portion. This is known to prone to bugs, due to the complexity
and number of the invariants assumed across netstack to maintain.
In the new PacketHeader API, client will call Push()/Consume() method to
construct/parse an outgoing/incoming packet. All invariants, such as slicing
and trimming, are maintained by the API itself.
NewPacketBuffer() is introduced to create new PacketBuffer. Zero value is no
longer valid.
PacketBuffer now assumes the packet is a concatenation of following portions:
* LinkHeader
* NetworkHeader
* TransportHeader
* Data
Any of them could be empty, or zero-length.
PiperOrigin-RevId: 326507688
TCP now tracks the overhead of the segment structure itself in it's out-of-order
queue (pending). This is required to ensure that a malicious sender sending 1
byte out-of-order segments cannot queue like 1000's of segments which bloat up
memory usage.
We also reduce the default receive window to 32KB. With TCP moderation there is
no need to keep this window at 1MB which means that for new connections the
default out-of-order queue will be small unless the application actually reads
the data that is being sent. This prevents a sender from just maliciously
filling up pending buf with lots of tiny out-of-order segments.
PiperOrigin-RevId: 323450913
Netstack has traditionally parsed headers on-demand as a packet moves up the
stack. This is conceptually simple and convenient, but incompatible with
iptables, where headers can be inspected and mangled before even a routing
decision is made.
This changes header parsing to happen early in the incoming packet path, as soon
as the NIC gets the packet from a link endpoint. Even if an invalid packet is
found (e.g. a TCP header of insufficient length), the packet is passed up the
stack for proper stats bookkeeping.
PiperOrigin-RevId: 315179302
Historically we've been passing PacketBuffer by shallow copying through out
the stack. Right now, this is only correct as the caller would not use
PacketBuffer after passing into the next layer in netstack.
With new buffer management effort in gVisor/netstack, PacketBuffer will
own a Buffer (to be added). Internally, both PacketBuffer and Buffer may
have pointers and shallow copying shouldn't be used.
Updates #2404.
PiperOrigin-RevId: 314610879
Every call to sender.NextSeg does not need to iterate from the
front of the writeList as in a given recovery episode we can cache
the last nextSeg returned. There cannot be a lower sequenced segment
that matches the next call to NextSeg as otherwise we would have
returned that instead in the previous call.
This fixes the issue of excessive CPU usage w/ large send buffers
where we spend a lot of time iterating from the front of the list on
every NextSeg invocation.
Further the following other bugs were also fixed:
* Iteration of segments never sent in NextSeg() when looking for segments for
retransmission that match step1/3/4 of the NextSeg algorithm
* Correctly setting rescueRxt only if the rescue segment was actually sent.
* Correctly initializing rescueRxt/highRxt when entering SACK recovery.
* Correctly re-arming the timer only on retransmissions when SACK is in use
and not for every segment being sent as it was being done before.
* Copy over xmitTime and xmitCount on segment clone.
* Move writeNext along when skipping over SACKED segments. This is required
to prevent spurious retransmissions where we end up retransmitting data
that was never lost.
PiperOrigin-RevId: 310387671
These methods let users eaily break the VectorisedView abstraction, and
allowed netstack to slip into pseudo-enforcement of the "all headers are
in the first View" invariant. Removing them and replacing with PullUp(n)
breaks this reliance and will make it easier to add iptables support and
rework network buffer management.
The new View.PullUp(n) method is low cost in the common case, when when
all the headers fit in the first View.
PiperOrigin-RevId: 308163542
These methods let users eaily break the VectorisedView abstraction, and
allowed netstack to slip into pseudo-enforcement of the "all headers are
in the first View" invariant. Removing them and replacing with PullUp(n)
breaks this reliance and will make it easier to add iptables support and
rework network buffer management.
The new View.PullUp(n) method is low cost in the common case, when when
all the headers fit in the first View.
Software GSO implementation currently has a complicated code path with
implicit assumptions that all packets to WritePackets carry same Data
and it does this to avoid allocations on the path etc. But this makes it
hard to reuse the WritePackets API.
This change breaks all such assumptions by introducing a new Vectorised
View API ReadToVV which can be used to cleanly split a VV into multiple
independent VVs. Further this change also makes packet buffers linkable
to form an intrusive list. This allows us to get rid of the array of
packet buffers that are passed in the WritePackets API call and replace
it with a list of packet buffers.
While this code does introduce some more allocations in the benchmarks
it doesn't cause any degradation.
Updates #231
PiperOrigin-RevId: 304731742
This is a precursor to be being able to build an intrusive list
of PacketBuffers for use in queuing disciplines being implemented.
Updates #2214
PiperOrigin-RevId: 302677662
This change better follows what is outlined in RFC 793 section 3.4 figure 12
where a listening socket should not accept a SYN-ACK segment in response to a
(potentially) old SYN segment.
Tests: Test that checks the TCP RST segment sent in response to a TCP SYN-ACK
segment received on a listening TCP endpoint.
PiperOrigin-RevId: 278893114
Based on the guidelines at
https://opensource.google.com/docs/releasing/authors/.
1. $ rg -l "Google LLC" | xargs sed -i 's/Google LLC.*/The gVisor Authors./'
2. Manual fixup of "Google Inc" references.
3. Add AUTHORS file. Authors may request to be added to this file.
4. Point netstack AUTHORS to gVisor AUTHORS. Drop CONTRIBUTORS.
Fixes#209
PiperOrigin-RevId: 245823212
Change-Id: I64530b24ad021a7d683137459cafc510f5ee1de9
This change does not make use of SACK information but adds support to track
SACK information and store it in the endpoint.
The actual SACK based recovery will be in a separate CL.
Part of commits to add RFC 6675 support to Netstack.
PiperOrigin-RevId: 235612264
Change-Id: I261f94844d7bad5abda803152ce6cc6125a467ff
So that when saving TCP endpoint in these states, there is no pending or
background activities.
Also lift tcp network save rejection error to tcpip package.
PiperOrigin-RevId: 199370748
Change-Id: Ief7b45c2a7338d12414cd7c23db95de6a9c22700