diff options
Diffstat (limited to 'pkg/tcpip/stack')
-rw-r--r-- | pkg/tcpip/stack/forwarder_test.go | 2 | ||||
-rw-r--r-- | pkg/tcpip/stack/iptables.go | 47 | ||||
-rw-r--r-- | pkg/tcpip/stack/iptables_types.go | 62 | ||||
-rw-r--r-- | pkg/tcpip/stack/nic.go | 9 | ||||
-rw-r--r-- | pkg/tcpip/stack/nic_test.go | 2 | ||||
-rw-r--r-- | pkg/tcpip/stack/registration.go | 21 | ||||
-rw-r--r-- | pkg/tcpip/stack/route.go | 15 | ||||
-rw-r--r-- | pkg/tcpip/stack/stack.go | 4 |
8 files changed, 96 insertions, 66 deletions
diff --git a/pkg/tcpip/stack/forwarder_test.go b/pkg/tcpip/stack/forwarder_test.go index 8084d50bc..344d60baa 100644 --- a/pkg/tcpip/stack/forwarder_test.go +++ b/pkg/tcpip/stack/forwarder_test.go @@ -209,7 +209,7 @@ func (e *fwdTestLinkEndpoint) InjectInbound(protocol tcpip.NetworkProtocolNumber // InjectLinkAddr injects an inbound packet with a remote link address. func (e *fwdTestLinkEndpoint) InjectLinkAddr(protocol tcpip.NetworkProtocolNumber, remote tcpip.LinkAddress, pkt PacketBuffer) { - e.dispatcher.DeliverNetworkPacket(e, remote, "" /* local */, protocol, pkt) + e.dispatcher.DeliverNetworkPacket(remote, "" /* local */, protocol, pkt) } // Attach saves the stack network-layer dispatcher for use later when packets diff --git a/pkg/tcpip/stack/iptables.go b/pkg/tcpip/stack/iptables.go index 443423b3c..709ede3fa 100644 --- a/pkg/tcpip/stack/iptables.go +++ b/pkg/tcpip/stack/iptables.go @@ -16,7 +16,6 @@ package stack import ( "fmt" - "strings" "gvisor.dev/gvisor/pkg/tcpip" "gvisor.dev/gvisor/pkg/tcpip/header" @@ -314,7 +313,7 @@ func (it *IPTables) checkRule(hook Hook, pkt *PacketBuffer, table Table, ruleIdx } // Check whether the packet matches the IP header filter. - if !filterMatch(rule.Filter, header.IPv4(pkt.NetworkHeader), hook, nicName) { + if !rule.Filter.match(header.IPv4(pkt.NetworkHeader), hook, nicName) { // Continue on to the next rule. return RuleJump, ruleIdx + 1 } @@ -335,47 +334,3 @@ func (it *IPTables) checkRule(hook Hook, pkt *PacketBuffer, table Table, ruleIdx // All the matchers matched, so run the target. return rule.Target.Action(pkt, &it.connections, hook, gso, r, address) } - -func filterMatch(filter IPHeaderFilter, hdr header.IPv4, hook Hook, nicName string) bool { - // TODO(gvisor.dev/issue/170): Support other fields of the filter. - // Check the transport protocol. - if filter.Protocol != 0 && filter.Protocol != hdr.TransportProtocol() { - return false - } - - // Check the destination IP. - dest := hdr.DestinationAddress() - matches := true - for i := range filter.Dst { - if dest[i]&filter.DstMask[i] != filter.Dst[i] { - matches = false - break - } - } - if matches == filter.DstInvert { - return false - } - - // Check the output interface. - // TODO(gvisor.dev/issue/170): Add the check for FORWARD and POSTROUTING - // hooks after supported. - if hook == Output { - n := len(filter.OutputInterface) - if n == 0 { - return true - } - - // If the interface name ends with '+', any interface which begins - // with the name should be matched. - ifName := filter.OutputInterface - matches = true - if strings.HasSuffix(ifName, "+") { - matches = strings.HasPrefix(nicName, ifName[:n-1]) - } else { - matches = nicName == ifName - } - return filter.OutputInterfaceInvert != matches - } - - return true -} diff --git a/pkg/tcpip/stack/iptables_types.go b/pkg/tcpip/stack/iptables_types.go index fe06007ae..a3bd3e700 100644 --- a/pkg/tcpip/stack/iptables_types.go +++ b/pkg/tcpip/stack/iptables_types.go @@ -15,7 +15,10 @@ package stack import ( + "strings" + "gvisor.dev/gvisor/pkg/tcpip" + "gvisor.dev/gvisor/pkg/tcpip/header" ) // A Hook specifies one of the hooks built into the network stack. @@ -159,6 +162,16 @@ type IPHeaderFilter struct { // comparison. DstInvert bool + // Src matches the source IP address. + Src tcpip.Address + + // SrcMask masks bits of the source IP address when comparing with Src. + SrcMask tcpip.Address + + // SrcInvert inverts the meaning of the source IP check, i.e. when true the + // filter will match packets that fail the source comparison. + SrcInvert bool + // OutputInterface matches the name of the outgoing interface for the // packet. OutputInterface string @@ -173,6 +186,55 @@ type IPHeaderFilter struct { OutputInterfaceInvert bool } +// match returns whether hdr matches the filter. +func (fl IPHeaderFilter) match(hdr header.IPv4, hook Hook, nicName string) bool { + // TODO(gvisor.dev/issue/170): Support other fields of the filter. + // Check the transport protocol. + if fl.Protocol != 0 && fl.Protocol != hdr.TransportProtocol() { + return false + } + + // Check the source and destination IPs. + if !filterAddress(hdr.DestinationAddress(), fl.DstMask, fl.Dst, fl.DstInvert) || !filterAddress(hdr.SourceAddress(), fl.SrcMask, fl.Src, fl.SrcInvert) { + return false + } + + // Check the output interface. + // TODO(gvisor.dev/issue/170): Add the check for FORWARD and POSTROUTING + // hooks after supported. + if hook == Output { + n := len(fl.OutputInterface) + if n == 0 { + return true + } + + // If the interface name ends with '+', any interface which begins + // with the name should be matched. + ifName := fl.OutputInterface + matches := true + if strings.HasSuffix(ifName, "+") { + matches = strings.HasPrefix(nicName, ifName[:n-1]) + } else { + matches = nicName == ifName + } + return fl.OutputInterfaceInvert != matches + } + + return true +} + +// filterAddress returns whether addr matches the filter. +func filterAddress(addr, mask, filterAddr tcpip.Address, invert bool) bool { + matches := true + for i := range filterAddr { + if addr[i]&mask[i] != filterAddr[i] { + matches = false + break + } + } + return matches != invert +} + // A Matcher is the interface for matching packets. type Matcher interface { // Name returns the name of the Matcher. diff --git a/pkg/tcpip/stack/nic.go b/pkg/tcpip/stack/nic.go index 54103fdb3..05646e5e2 100644 --- a/pkg/tcpip/stack/nic.go +++ b/pkg/tcpip/stack/nic.go @@ -1167,7 +1167,7 @@ func handlePacket(protocol tcpip.NetworkProtocolNumber, dst, src tcpip.Address, // Note that the ownership of the slice backing vv is retained by the caller. // This rule applies only to the slice itself, not to the items of the slice; // the ownership of the items is not retained by the caller. -func (n *NIC) DeliverNetworkPacket(linkEP LinkEndpoint, remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt PacketBuffer) { +func (n *NIC) DeliverNetworkPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt PacketBuffer) { n.mu.RLock() enabled := n.mu.enabled // If the NIC is not yet enabled, don't receive any packets. @@ -1240,7 +1240,7 @@ func (n *NIC) DeliverNetworkPacket(linkEP LinkEndpoint, remote, local tcpip.Link } if ref := n.getRef(protocol, dst); ref != nil { - handlePacket(protocol, dst, src, linkEP.LinkAddress(), remote, ref, pkt) + handlePacket(protocol, dst, src, n.linkEP.LinkAddress(), remote, ref, pkt) return } @@ -1304,13 +1304,16 @@ func (n *NIC) forwardPacket(r *Route, protocol tcpip.NetworkProtocolNumber, pkt pkt.Header = buffer.NewPrependable(linkHeaderLen) } + // WritePacket takes ownership of pkt, calculate numBytes first. + numBytes := pkt.Header.UsedLength() + pkt.Data.Size() + if err := n.linkEP.WritePacket(r, nil /* gso */, protocol, pkt); err != nil { r.Stats().IP.OutgoingPacketErrors.Increment() return } n.stats.Tx.Packets.Increment() - n.stats.Tx.Bytes.IncrementBy(uint64(pkt.Header.UsedLength() + pkt.Data.Size())) + n.stats.Tx.Bytes.IncrementBy(uint64(numBytes)) } // DeliverTransportPacket delivers the packets to the appropriate transport diff --git a/pkg/tcpip/stack/nic_test.go b/pkg/tcpip/stack/nic_test.go index d672fc157..b01b3f476 100644 --- a/pkg/tcpip/stack/nic_test.go +++ b/pkg/tcpip/stack/nic_test.go @@ -44,7 +44,7 @@ func TestDisabledRxStatsWhenNICDisabled(t *testing.T) { t.FailNow() } - nic.DeliverNetworkPacket(nil, "", "", 0, PacketBuffer{Data: buffer.View([]byte{1, 2, 3, 4}).ToVectorisedView()}) + nic.DeliverNetworkPacket("", "", 0, PacketBuffer{Data: buffer.View([]byte{1, 2, 3, 4}).ToVectorisedView()}) if got := nic.stats.DisabledRx.Packets.Value(); got != 1 { t.Errorf("got DisabledRx.Packets = %d, want = 1", got) diff --git a/pkg/tcpip/stack/registration.go b/pkg/tcpip/stack/registration.go index b331427c6..db89234e8 100644 --- a/pkg/tcpip/stack/registration.go +++ b/pkg/tcpip/stack/registration.go @@ -240,16 +240,17 @@ type NetworkEndpoint interface { MaxHeaderLength() uint16 // WritePacket writes a packet to the given destination address and - // protocol. It sets pkt.NetworkHeader. pkt.TransportHeader must have - // already been set. + // protocol. It takes ownership of pkt. pkt.TransportHeader must have already + // been set. WritePacket(r *Route, gso *GSO, params NetworkHeaderParams, pkt PacketBuffer) *tcpip.Error // WritePackets writes packets to the given destination address and - // protocol. pkts must not be zero length. + // protocol. pkts must not be zero length. It takes ownership of pkts and + // underlying packets. WritePackets(r *Route, gso *GSO, pkts PacketBufferList, params NetworkHeaderParams) (int, *tcpip.Error) // WriteHeaderIncludedPacket writes a packet that includes a network - // header to the given destination address. + // header to the given destination address. It takes ownership of pkt. WriteHeaderIncludedPacket(r *Route, pkt PacketBuffer) *tcpip.Error // ID returns the network protocol endpoint ID. @@ -326,7 +327,7 @@ type NetworkDispatcher interface { // packets sent via loopback), and won't have the field set. // // DeliverNetworkPacket takes ownership of pkt. - DeliverNetworkPacket(linkEP LinkEndpoint, remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt PacketBuffer) + DeliverNetworkPacket(remote, local tcpip.LinkAddress, protocol tcpip.NetworkProtocolNumber, pkt PacketBuffer) } // LinkEndpointCapabilities is the type associated with the capabilities @@ -382,9 +383,8 @@ type LinkEndpoint interface { LinkAddress() tcpip.LinkAddress // WritePacket writes a packet with the given protocol through the - // given route. It sets pkt.LinkHeader if a link layer header exists. - // pkt.NetworkHeader and pkt.TransportHeader must have already been - // set. + // given route. It takes ownership of pkt. pkt.NetworkHeader and + // pkt.TransportHeader must have already been set. // // To participate in transparent bridging, a LinkEndpoint implementation // should call eth.Encode with header.EthernetFields.SrcAddr set to @@ -392,7 +392,8 @@ type LinkEndpoint interface { WritePacket(r *Route, gso *GSO, protocol tcpip.NetworkProtocolNumber, pkt PacketBuffer) *tcpip.Error // WritePackets writes packets with the given protocol through the - // given route. pkts must not be zero length. + // given route. pkts must not be zero length. It takes ownership of pkts and + // underlying packets. // // Right now, WritePackets is used only when the software segmentation // offload is enabled. If it will be used for something else, it may @@ -400,7 +401,7 @@ type LinkEndpoint interface { WritePackets(r *Route, gso *GSO, pkts PacketBufferList, protocol tcpip.NetworkProtocolNumber) (int, *tcpip.Error) // WriteRawPacket writes a packet directly to the link. The packet - // should already have an ethernet header. + // should already have an ethernet header. It takes ownership of vv. WriteRawPacket(vv buffer.VectorisedView) *tcpip.Error // Attach attaches the data link layer endpoint to the network-layer diff --git a/pkg/tcpip/stack/route.go b/pkg/tcpip/stack/route.go index 150297ab9..3d0e5cc6e 100644 --- a/pkg/tcpip/stack/route.go +++ b/pkg/tcpip/stack/route.go @@ -158,12 +158,15 @@ func (r *Route) WritePacket(gso *GSO, params NetworkHeaderParams, pkt PacketBuff return tcpip.ErrInvalidEndpointState } + // WritePacket takes ownership of pkt, calculate numBytes first. + numBytes := pkt.Header.UsedLength() + pkt.Data.Size() + err := r.ref.ep.WritePacket(r, gso, params, pkt) if err != nil { r.Stats().IP.OutgoingPacketErrors.Increment() } else { r.ref.nic.stats.Tx.Packets.Increment() - r.ref.nic.stats.Tx.Bytes.IncrementBy(uint64(pkt.Header.UsedLength() + pkt.Data.Size())) + r.ref.nic.stats.Tx.Bytes.IncrementBy(uint64(numBytes)) } return err } @@ -175,9 +178,12 @@ func (r *Route) WritePackets(gso *GSO, pkts PacketBufferList, params NetworkHead return 0, tcpip.ErrInvalidEndpointState } + // WritePackets takes ownership of pkt, calculate length first. + numPkts := pkts.Len() + n, err := r.ref.ep.WritePackets(r, gso, pkts, params) if err != nil { - r.Stats().IP.OutgoingPacketErrors.IncrementBy(uint64(pkts.Len() - n)) + r.Stats().IP.OutgoingPacketErrors.IncrementBy(uint64(numPkts - n)) } r.ref.nic.stats.Tx.Packets.IncrementBy(uint64(n)) @@ -198,12 +204,15 @@ func (r *Route) WriteHeaderIncludedPacket(pkt PacketBuffer) *tcpip.Error { return tcpip.ErrInvalidEndpointState } + // WriteHeaderIncludedPacket takes ownership of pkt, calculate numBytes first. + numBytes := pkt.Data.Size() + if err := r.ref.ep.WriteHeaderIncludedPacket(r, pkt); err != nil { r.Stats().IP.OutgoingPacketErrors.Increment() return err } r.ref.nic.stats.Tx.Packets.Increment() - r.ref.nic.stats.Tx.Bytes.IncrementBy(uint64(pkt.Data.Size())) + r.ref.nic.stats.Tx.Bytes.IncrementBy(uint64(numBytes)) return nil } diff --git a/pkg/tcpip/stack/stack.go b/pkg/tcpip/stack/stack.go index b39ffa9fb..0ab4c3e19 100644 --- a/pkg/tcpip/stack/stack.go +++ b/pkg/tcpip/stack/stack.go @@ -235,11 +235,11 @@ type RcvBufAutoTuneParams struct { // was started. MeasureTime time.Time - // CopiedBytes is the number of bytes copied to user space since + // CopiedBytes is the number of bytes copied to userspace since // this measure began. CopiedBytes int - // PrevCopiedBytes is the number of bytes copied to user space in + // PrevCopiedBytes is the number of bytes copied to userspace in // the previous RTT period. PrevCopiedBytes int |