diff options
Diffstat (limited to 'pkg/sentry/socket')
-rw-r--r-- | pkg/sentry/socket/hostinet/BUILD | 1 | ||||
-rw-r--r-- | pkg/sentry/socket/hostinet/socket.go | 2 | ||||
-rw-r--r-- | pkg/sentry/socket/hostinet/socket_vfs2.go | 3 | ||||
-rw-r--r-- | pkg/sentry/socket/netfilter/netfilter.go | 57 | ||||
-rw-r--r-- | pkg/sentry/socket/netfilter/owner_matcher.go | 2 | ||||
-rw-r--r-- | pkg/sentry/socket/netfilter/tcp_matcher.go | 36 | ||||
-rw-r--r-- | pkg/sentry/socket/netfilter/udp_matcher.go | 36 | ||||
-rw-r--r-- | pkg/sentry/socket/netlink/BUILD | 1 | ||||
-rw-r--r-- | pkg/sentry/socket/netlink/socket_vfs2.go | 8 | ||||
-rw-r--r-- | pkg/sentry/socket/netstack/BUILD | 2 | ||||
-rw-r--r-- | pkg/sentry/socket/netstack/netstack.go | 50 | ||||
-rw-r--r-- | pkg/sentry/socket/netstack/netstack_vfs2.go | 3 | ||||
-rw-r--r-- | pkg/sentry/socket/netstack/stack.go | 9 | ||||
-rw-r--r-- | pkg/sentry/socket/unix/BUILD | 1 | ||||
-rw-r--r-- | pkg/sentry/socket/unix/transport/connectioned.go | 2 | ||||
-rw-r--r-- | pkg/sentry/socket/unix/unix.go | 39 | ||||
-rw-r--r-- | pkg/sentry/socket/unix/unix_vfs2.go | 7 |
17 files changed, 136 insertions, 123 deletions
diff --git a/pkg/sentry/socket/hostinet/BUILD b/pkg/sentry/socket/hostinet/BUILD index e82d6cd1e..60c9896fc 100644 --- a/pkg/sentry/socket/hostinet/BUILD +++ b/pkg/sentry/socket/hostinet/BUILD @@ -34,6 +34,7 @@ go_library( "//pkg/sentry/socket", "//pkg/sentry/socket/control", "//pkg/sentry/vfs", + "//pkg/sentry/vfs/lock", "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip/stack", diff --git a/pkg/sentry/socket/hostinet/socket.go b/pkg/sentry/socket/hostinet/socket.go index b49433326..c11e82c10 100644 --- a/pkg/sentry/socket/hostinet/socket.go +++ b/pkg/sentry/socket/hostinet/socket.go @@ -555,7 +555,7 @@ func (s *socketOpsCommon) SendMsg(t *kernel.Task, src usermem.IOSequence, to []b if uint64(src.NumBytes()) != srcs.NumBytes() { return 0, nil } - if srcs.IsEmpty() { + if srcs.IsEmpty() && len(controlBuf) == 0 { return 0, nil } diff --git a/pkg/sentry/socket/hostinet/socket_vfs2.go b/pkg/sentry/socket/hostinet/socket_vfs2.go index 677743113..027add1fd 100644 --- a/pkg/sentry/socket/hostinet/socket_vfs2.go +++ b/pkg/sentry/socket/hostinet/socket_vfs2.go @@ -26,6 +26,7 @@ import ( "gvisor.dev/gvisor/pkg/sentry/kernel" "gvisor.dev/gvisor/pkg/sentry/socket" "gvisor.dev/gvisor/pkg/sentry/vfs" + "gvisor.dev/gvisor/pkg/sentry/vfs/lock" "gvisor.dev/gvisor/pkg/syserr" "gvisor.dev/gvisor/pkg/syserror" "gvisor.dev/gvisor/pkg/usermem" @@ -35,6 +36,7 @@ import ( type socketVFS2 struct { vfsfd vfs.FileDescription vfs.FileDescriptionDefaultImpl + vfs.LockFD // We store metadata for hostinet sockets internally. Technically, we should // access metadata (e.g. through stat, chmod) on the host for correctness, @@ -59,6 +61,7 @@ func newVFS2Socket(t *kernel.Task, family int, stype linux.SockType, protocol in fd: fd, }, } + s.LockFD.Init(&lock.FileLocks{}) if err := fdnotifier.AddFD(int32(fd), &s.queue); err != nil { return nil, syserr.FromError(err) } diff --git a/pkg/sentry/socket/netfilter/netfilter.go b/pkg/sentry/socket/netfilter/netfilter.go index 789bb94c8..66015e2bc 100644 --- a/pkg/sentry/socket/netfilter/netfilter.go +++ b/pkg/sentry/socket/netfilter/netfilter.go @@ -64,6 +64,8 @@ const enableLogging = false var emptyFilter = stack.IPHeaderFilter{ Dst: "\x00\x00\x00\x00", DstMask: "\x00\x00\x00\x00", + Src: "\x00\x00\x00\x00", + SrcMask: "\x00\x00\x00\x00", } // nflog logs messages related to the writing and reading of iptables. @@ -142,31 +144,27 @@ func GetEntries(t *kernel.Task, stack *stack.Stack, outPtr usermem.Addr, outLen } func findTable(stk *stack.Stack, tablename linux.TableName) (stack.Table, error) { - ipt := stk.IPTables() - table, ok := ipt.Tables[tablename.String()] + table, ok := stk.IPTables().GetTable(tablename.String()) if !ok { return stack.Table{}, fmt.Errorf("couldn't find table %q", tablename) } return table, nil } -// FillDefaultIPTables sets stack's IPTables to the default tables and -// populates them with metadata. -func FillDefaultIPTables(stk *stack.Stack) { - ipt := stack.DefaultTables() - - // In order to fill in the metadata, we have to translate ipt from its - // netstack format to Linux's giant-binary-blob format. - for name, table := range ipt.Tables { - _, metadata, err := convertNetstackToBinary(name, table) - if err != nil { - panic(fmt.Errorf("Unable to set default IP tables: %v", err)) +// FillIPTablesMetadata populates stack's IPTables with metadata. +func FillIPTablesMetadata(stk *stack.Stack) { + stk.IPTables().ModifyTables(func(tables map[string]stack.Table) { + // In order to fill in the metadata, we have to translate ipt from its + // netstack format to Linux's giant-binary-blob format. + for name, table := range tables { + _, metadata, err := convertNetstackToBinary(name, table) + if err != nil { + panic(fmt.Errorf("Unable to set default IP tables: %v", err)) + } + table.SetMetadata(metadata) + tables[name] = table } - table.SetMetadata(metadata) - ipt.Tables[name] = table - } - - stk.SetIPTables(ipt) + }) } // convertNetstackToBinary converts the iptables as stored in netstack to the @@ -214,11 +212,16 @@ func convertNetstackToBinary(tablename string, table stack.Table) (linux.KernelI } copy(entry.IPTEntry.IP.Dst[:], rule.Filter.Dst) copy(entry.IPTEntry.IP.DstMask[:], rule.Filter.DstMask) + copy(entry.IPTEntry.IP.Src[:], rule.Filter.Src) + copy(entry.IPTEntry.IP.SrcMask[:], rule.Filter.SrcMask) copy(entry.IPTEntry.IP.OutputInterface[:], rule.Filter.OutputInterface) copy(entry.IPTEntry.IP.OutputInterfaceMask[:], rule.Filter.OutputInterfaceMask) if rule.Filter.DstInvert { entry.IPTEntry.IP.InverseFlags |= linux.IPT_INV_DSTIP } + if rule.Filter.SrcInvert { + entry.IPTEntry.IP.InverseFlags |= linux.IPT_INV_SRCIP + } if rule.Filter.OutputInterfaceInvert { entry.IPTEntry.IP.InverseFlags |= linux.IPT_INV_VIA_OUT } @@ -566,15 +569,13 @@ func SetEntries(stk *stack.Stack, optVal []byte) *syserr.Error { // - There are no chains without an unconditional final rule. // - There are no chains without an unconditional underflow rule. - ipt := stk.IPTables() table.SetMetadata(metadata{ HookEntry: replace.HookEntry, Underflow: replace.Underflow, NumEntries: replace.NumEntries, Size: replace.Size, }) - ipt.Tables[replace.Name.String()] = table - stk.SetIPTables(ipt) + stk.IPTables().ReplaceTable(replace.Name.String(), table) return nil } @@ -737,6 +738,9 @@ func filterFromIPTIP(iptip linux.IPTIP) (stack.IPHeaderFilter, error) { if len(iptip.Dst) != header.IPv4AddressSize || len(iptip.DstMask) != header.IPv4AddressSize { return stack.IPHeaderFilter{}, fmt.Errorf("incorrect length of destination (%d) and/or destination mask (%d) fields", len(iptip.Dst), len(iptip.DstMask)) } + if len(iptip.Src) != header.IPv4AddressSize || len(iptip.SrcMask) != header.IPv4AddressSize { + return stack.IPHeaderFilter{}, fmt.Errorf("incorrect length of source (%d) and/or source mask (%d) fields", len(iptip.Src), len(iptip.SrcMask)) + } n := bytes.IndexByte([]byte(iptip.OutputInterface[:]), 0) if n == -1 { @@ -755,6 +759,9 @@ func filterFromIPTIP(iptip linux.IPTIP) (stack.IPHeaderFilter, error) { Dst: tcpip.Address(iptip.Dst[:]), DstMask: tcpip.Address(iptip.DstMask[:]), DstInvert: iptip.InverseFlags&linux.IPT_INV_DSTIP != 0, + Src: tcpip.Address(iptip.Src[:]), + SrcMask: tcpip.Address(iptip.SrcMask[:]), + SrcInvert: iptip.InverseFlags&linux.IPT_INV_SRCIP != 0, OutputInterface: ifname, OutputInterfaceMask: ifnameMask, OutputInterfaceInvert: iptip.InverseFlags&linux.IPT_INV_VIA_OUT != 0, @@ -765,15 +772,13 @@ func containsUnsupportedFields(iptip linux.IPTIP) bool { // The following features are supported: // - Protocol // - Dst and DstMask + // - Src and SrcMask // - The inverse destination IP check flag // - OutputInterface, OutputInterfaceMask and its inverse. - var emptyInetAddr = linux.InetAddr{} var emptyInterface = [linux.IFNAMSIZ]byte{} // Disable any supported inverse flags. - inverseMask := uint8(linux.IPT_INV_DSTIP) | uint8(linux.IPT_INV_VIA_OUT) - return iptip.Src != emptyInetAddr || - iptip.SrcMask != emptyInetAddr || - iptip.InputInterface != emptyInterface || + inverseMask := uint8(linux.IPT_INV_DSTIP) | uint8(linux.IPT_INV_SRCIP) | uint8(linux.IPT_INV_VIA_OUT) + return iptip.InputInterface != emptyInterface || iptip.InputInterfaceMask != emptyInterface || iptip.Flags != 0 || iptip.InverseFlags&^inverseMask != 0 diff --git a/pkg/sentry/socket/netfilter/owner_matcher.go b/pkg/sentry/socket/netfilter/owner_matcher.go index 3863293c7..1b4e0ad79 100644 --- a/pkg/sentry/socket/netfilter/owner_matcher.go +++ b/pkg/sentry/socket/netfilter/owner_matcher.go @@ -111,7 +111,7 @@ func (*OwnerMatcher) Name() string { } // Match implements Matcher.Match. -func (om *OwnerMatcher) Match(hook stack.Hook, pkt stack.PacketBuffer, interfaceName string) (bool, bool) { +func (om *OwnerMatcher) Match(hook stack.Hook, pkt *stack.PacketBuffer, interfaceName string) (bool, bool) { // Support only for OUTPUT chain. // TODO(gvisor.dev/issue/170): Need to support for POSTROUTING chain also. if hook != stack.Output { diff --git a/pkg/sentry/socket/netfilter/tcp_matcher.go b/pkg/sentry/socket/netfilter/tcp_matcher.go index 57a1e1c12..4f98ee2d5 100644 --- a/pkg/sentry/socket/netfilter/tcp_matcher.go +++ b/pkg/sentry/socket/netfilter/tcp_matcher.go @@ -96,7 +96,7 @@ func (*TCPMatcher) Name() string { } // Match implements Matcher.Match. -func (tm *TCPMatcher) Match(hook stack.Hook, pkt stack.PacketBuffer, interfaceName string) (bool, bool) { +func (tm *TCPMatcher) Match(hook stack.Hook, pkt *stack.PacketBuffer, interfaceName string) (bool, bool) { netHeader := header.IPv4(pkt.NetworkHeader) if netHeader.TransportProtocol() != header.TCPProtocolNumber { @@ -111,36 +111,10 @@ func (tm *TCPMatcher) Match(hook stack.Hook, pkt stack.PacketBuffer, interfaceNa return false, false } - // Now we need the transport header. However, this may not have been set - // yet. - // TODO(gvisor.dev/issue/170): Parsing the transport header should - // ultimately be moved into the stack.Check codepath as matchers are - // added. - var tcpHeader header.TCP - if pkt.TransportHeader != nil { - tcpHeader = header.TCP(pkt.TransportHeader) - } else { - var length int - if hook == stack.Prerouting { - // The network header hasn't been parsed yet. We have to do it here. - hdr, ok := pkt.Data.PullUp(header.IPv4MinimumSize) - if !ok { - // There's no valid TCP header here, so we hotdrop the - // packet. - return false, true - } - h := header.IPv4(hdr) - pkt.NetworkHeader = hdr - length = int(h.HeaderLength()) - } - // The TCP header hasn't been parsed yet. We have to do it here. - hdr, ok := pkt.Data.PullUp(length + header.TCPMinimumSize) - if !ok { - // There's no valid TCP header here, so we hotdrop the - // packet. - return false, true - } - tcpHeader = header.TCP(hdr[length:]) + tcpHeader := header.TCP(pkt.TransportHeader) + if len(tcpHeader) < header.TCPMinimumSize { + // There's no valid TCP header here, so we drop the packet immediately. + return false, true } // Check whether the source and destination ports are within the diff --git a/pkg/sentry/socket/netfilter/udp_matcher.go b/pkg/sentry/socket/netfilter/udp_matcher.go index cfa9e621d..3f20fc891 100644 --- a/pkg/sentry/socket/netfilter/udp_matcher.go +++ b/pkg/sentry/socket/netfilter/udp_matcher.go @@ -93,7 +93,7 @@ func (*UDPMatcher) Name() string { } // Match implements Matcher.Match. -func (um *UDPMatcher) Match(hook stack.Hook, pkt stack.PacketBuffer, interfaceName string) (bool, bool) { +func (um *UDPMatcher) Match(hook stack.Hook, pkt *stack.PacketBuffer, interfaceName string) (bool, bool) { netHeader := header.IPv4(pkt.NetworkHeader) // TODO(gvisor.dev/issue/170): Proto checks should ultimately be moved @@ -110,36 +110,10 @@ func (um *UDPMatcher) Match(hook stack.Hook, pkt stack.PacketBuffer, interfaceNa return false, false } - // Now we need the transport header. However, this may not have been set - // yet. - // TODO(gvisor.dev/issue/170): Parsing the transport header should - // ultimately be moved into the stack.Check codepath as matchers are - // added. - var udpHeader header.UDP - if pkt.TransportHeader != nil { - udpHeader = header.UDP(pkt.TransportHeader) - } else { - var length int - if hook == stack.Prerouting { - // The network header hasn't been parsed yet. We have to do it here. - hdr, ok := pkt.Data.PullUp(header.IPv4MinimumSize) - if !ok { - // There's no valid UDP header here, so we hotdrop the - // packet. - return false, true - } - h := header.IPv4(hdr) - pkt.NetworkHeader = hdr - length = int(h.HeaderLength()) - } - // The UDP header hasn't been parsed yet. We have to do it here. - hdr, ok := pkt.Data.PullUp(length + header.UDPMinimumSize) - if !ok { - // There's no valid UDP header here, so we hotdrop the - // packet. - return false, true - } - udpHeader = header.UDP(hdr[length:]) + udpHeader := header.UDP(pkt.TransportHeader) + if len(udpHeader) < header.UDPMinimumSize { + // There's no valid UDP header here, so we drop the packet immediately. + return false, true } // Check whether the source and destination ports are within the diff --git a/pkg/sentry/socket/netlink/BUILD b/pkg/sentry/socket/netlink/BUILD index 7212d8644..420e573c9 100644 --- a/pkg/sentry/socket/netlink/BUILD +++ b/pkg/sentry/socket/netlink/BUILD @@ -29,6 +29,7 @@ go_library( "//pkg/sentry/socket/unix", "//pkg/sentry/socket/unix/transport", "//pkg/sentry/vfs", + "//pkg/sentry/vfs/lock", "//pkg/sync", "//pkg/syserr", "//pkg/syserror", diff --git a/pkg/sentry/socket/netlink/socket_vfs2.go b/pkg/sentry/socket/netlink/socket_vfs2.go index b854bf990..8bfee5193 100644 --- a/pkg/sentry/socket/netlink/socket_vfs2.go +++ b/pkg/sentry/socket/netlink/socket_vfs2.go @@ -23,6 +23,7 @@ import ( "gvisor.dev/gvisor/pkg/sentry/socket/unix" "gvisor.dev/gvisor/pkg/sentry/socket/unix/transport" "gvisor.dev/gvisor/pkg/sentry/vfs" + "gvisor.dev/gvisor/pkg/sentry/vfs/lock" "gvisor.dev/gvisor/pkg/syserr" "gvisor.dev/gvisor/pkg/syserror" "gvisor.dev/gvisor/pkg/tcpip" @@ -40,6 +41,7 @@ type SocketVFS2 struct { vfsfd vfs.FileDescription vfs.FileDescriptionDefaultImpl vfs.DentryMetadataFileDescriptionImpl + vfs.LockFD socketOpsCommon } @@ -66,7 +68,7 @@ func NewVFS2(t *kernel.Task, skType linux.SockType, protocol Protocol) (*SocketV return nil, err } - return &SocketVFS2{ + fd := &SocketVFS2{ socketOpsCommon: socketOpsCommon{ ports: t.Kernel().NetlinkPorts(), protocol: protocol, @@ -75,7 +77,9 @@ func NewVFS2(t *kernel.Task, skType linux.SockType, protocol Protocol) (*SocketV connection: connection, sendBufferSize: defaultSendBufferSize, }, - }, nil + } + fd.LockFD.Init(&lock.FileLocks{}) + return fd, nil } // Readiness implements waiter.Waitable.Readiness. diff --git a/pkg/sentry/socket/netstack/BUILD b/pkg/sentry/socket/netstack/BUILD index 333e0042e..0f592ecc3 100644 --- a/pkg/sentry/socket/netstack/BUILD +++ b/pkg/sentry/socket/netstack/BUILD @@ -37,6 +37,7 @@ go_library( "//pkg/sentry/socket/netfilter", "//pkg/sentry/unimpl", "//pkg/sentry/vfs", + "//pkg/sentry/vfs/lock", "//pkg/sync", "//pkg/syserr", "//pkg/syserror", @@ -50,5 +51,6 @@ go_library( "//pkg/tcpip/transport/udp", "//pkg/usermem", "//pkg/waiter", + "@org_golang_x_sys//unix:go_default_library", ], ) diff --git a/pkg/sentry/socket/netstack/netstack.go b/pkg/sentry/socket/netstack/netstack.go index 9dea2b5ff..738277391 100644 --- a/pkg/sentry/socket/netstack/netstack.go +++ b/pkg/sentry/socket/netstack/netstack.go @@ -33,6 +33,7 @@ import ( "syscall" "time" + "golang.org/x/sys/unix" "gvisor.dev/gvisor/pkg/abi/linux" "gvisor.dev/gvisor/pkg/amutex" "gvisor.dev/gvisor/pkg/binary" @@ -719,6 +720,14 @@ func (s *socketOpsCommon) Connect(t *kernel.Task, sockaddr []byte, blocking bool defer s.EventUnregister(&e) if err := s.Endpoint.Connect(addr); err != tcpip.ErrConnectStarted && err != tcpip.ErrAlreadyConnecting { + if (s.family == unix.AF_INET || s.family == unix.AF_INET6) && s.skType == linux.SOCK_STREAM { + // TCP unlike UDP returns EADDRNOTAVAIL when it can't + // find an available local ephemeral port. + if err == tcpip.ErrNoPortAvailable { + return syserr.ErrAddressNotAvailable + } + } + return syserr.TranslateNetstackError(err) } @@ -1237,6 +1246,18 @@ func getSockOptTCP(t *kernel.Task, ep commonEndpoint, name, outLen int) (interfa return int32(time.Duration(v) / time.Second), nil + case linux.TCP_KEEPCNT: + if outLen < sizeOfInt32 { + return nil, syserr.ErrInvalidArgument + } + + v, err := ep.GetSockOptInt(tcpip.KeepaliveCountOption) + if err != nil { + return nil, syserr.TranslateNetstackError(err) + } + + return int32(v), nil + case linux.TCP_USER_TIMEOUT: if outLen < sizeOfInt32 { return nil, syserr.ErrInvalidArgument @@ -1777,6 +1798,17 @@ func setSockOptTCP(t *kernel.Task, ep commonEndpoint, name int, optVal []byte) * } return syserr.TranslateNetstackError(ep.SetSockOpt(tcpip.KeepaliveIntervalOption(time.Second * time.Duration(v)))) + case linux.TCP_KEEPCNT: + if len(optVal) < sizeOfInt32 { + return syserr.ErrInvalidArgument + } + + v := usermem.ByteOrder.Uint32(optVal) + if v < 1 || v > linux.MAX_TCP_KEEPCNT { + return syserr.ErrInvalidArgument + } + return syserr.TranslateNetstackError(ep.SetSockOptInt(tcpip.KeepaliveCountOption, int(v))) + case linux.TCP_USER_TIMEOUT: if len(optVal) < sizeOfInt32 { return syserr.ErrInvalidArgument @@ -2106,30 +2138,20 @@ func emitUnimplementedEventTCP(t *kernel.Task, name int) { switch name { case linux.TCP_CONGESTION, linux.TCP_CORK, - linux.TCP_DEFER_ACCEPT, linux.TCP_FASTOPEN, linux.TCP_FASTOPEN_CONNECT, linux.TCP_FASTOPEN_KEY, linux.TCP_FASTOPEN_NO_COOKIE, - linux.TCP_KEEPCNT, - linux.TCP_KEEPIDLE, - linux.TCP_KEEPINTVL, - linux.TCP_LINGER2, - linux.TCP_MAXSEG, linux.TCP_QUEUE_SEQ, - linux.TCP_QUICKACK, linux.TCP_REPAIR, linux.TCP_REPAIR_QUEUE, linux.TCP_REPAIR_WINDOW, linux.TCP_SAVED_SYN, linux.TCP_SAVE_SYN, - linux.TCP_SYNCNT, linux.TCP_THIN_DUPACK, linux.TCP_THIN_LINEAR_TIMEOUTS, linux.TCP_TIMESTAMP, - linux.TCP_ULP, - linux.TCP_USER_TIMEOUT, - linux.TCP_WINDOW_CLAMP: + linux.TCP_ULP: t.Kernel().EmitUnimplementedEvent(t) } @@ -2718,7 +2740,7 @@ func (s *socketOpsCommon) ioctl(ctx context.Context, io usermem.IO, args arch.Sy v = math.MaxInt32 } - // Copy result to user-space. + // Copy result to userspace. _, err := usermem.CopyObjectOut(ctx, io, args[2].Pointer(), int32(v), usermem.IOOpts{ AddressSpaceActive: true, }) @@ -2787,7 +2809,7 @@ func Ioctl(ctx context.Context, ep commonEndpoint, io usermem.IO, args arch.Sysc if v > math.MaxInt32 { v = math.MaxInt32 } - // Copy result to user-space. + // Copy result to userspace. _, err := usermem.CopyObjectOut(ctx, io, args[2].Pointer(), int32(v), usermem.IOOpts{ AddressSpaceActive: true, }) @@ -2803,7 +2825,7 @@ func Ioctl(ctx context.Context, ep commonEndpoint, io usermem.IO, args arch.Sysc v = math.MaxInt32 } - // Copy result to user-space. + // Copy result to userspace. _, err := usermem.CopyObjectOut(ctx, io, args[2].Pointer(), int32(v), usermem.IOOpts{ AddressSpaceActive: true, }) diff --git a/pkg/sentry/socket/netstack/netstack_vfs2.go b/pkg/sentry/socket/netstack/netstack_vfs2.go index fcd8013c0..1412a4810 100644 --- a/pkg/sentry/socket/netstack/netstack_vfs2.go +++ b/pkg/sentry/socket/netstack/netstack_vfs2.go @@ -25,6 +25,7 @@ import ( "gvisor.dev/gvisor/pkg/sentry/socket" "gvisor.dev/gvisor/pkg/sentry/socket/netfilter" "gvisor.dev/gvisor/pkg/sentry/vfs" + "gvisor.dev/gvisor/pkg/sentry/vfs/lock" "gvisor.dev/gvisor/pkg/syserr" "gvisor.dev/gvisor/pkg/syserror" "gvisor.dev/gvisor/pkg/tcpip" @@ -38,6 +39,7 @@ type SocketVFS2 struct { vfsfd vfs.FileDescription vfs.FileDescriptionDefaultImpl vfs.DentryMetadataFileDescriptionImpl + vfs.LockFD socketOpsCommon } @@ -64,6 +66,7 @@ func NewVFS2(t *kernel.Task, family int, skType linux.SockType, protocol int, qu protocol: protocol, }, } + s.LockFD.Init(&lock.FileLocks{}) vfsfd := &s.vfsfd if err := vfsfd.Init(s, linux.O_RDWR, mnt, d, &vfs.FileDescriptionOptions{ DenyPRead: true, diff --git a/pkg/sentry/socket/netstack/stack.go b/pkg/sentry/socket/netstack/stack.go index f5fa18136..9b44c2b89 100644 --- a/pkg/sentry/socket/netstack/stack.go +++ b/pkg/sentry/socket/netstack/stack.go @@ -362,14 +362,13 @@ func (s *Stack) RouteTable() []inet.Route { } // IPTables returns the stack's iptables. -func (s *Stack) IPTables() (stack.IPTables, error) { +func (s *Stack) IPTables() (*stack.IPTables, error) { return s.Stack.IPTables(), nil } -// FillDefaultIPTables sets the stack's iptables to the default tables, which -// allow and do not modify all traffic. -func (s *Stack) FillDefaultIPTables() { - netfilter.FillDefaultIPTables(s.Stack) +// FillIPTablesMetadata populates stack's IPTables with metadata. +func (s *Stack) FillIPTablesMetadata() { + netfilter.FillIPTablesMetadata(s.Stack) } // Resume implements inet.Stack.Resume. diff --git a/pkg/sentry/socket/unix/BUILD b/pkg/sentry/socket/unix/BUILD index de2cc4bdf..7d4cc80fe 100644 --- a/pkg/sentry/socket/unix/BUILD +++ b/pkg/sentry/socket/unix/BUILD @@ -29,6 +29,7 @@ go_library( "//pkg/sentry/socket/netstack", "//pkg/sentry/socket/unix/transport", "//pkg/sentry/vfs", + "//pkg/sentry/vfs/lock", "//pkg/syserr", "//pkg/syserror", "//pkg/tcpip", diff --git a/pkg/sentry/socket/unix/transport/connectioned.go b/pkg/sentry/socket/unix/transport/connectioned.go index ce5b94ee7..09c6d3b27 100644 --- a/pkg/sentry/socket/unix/transport/connectioned.go +++ b/pkg/sentry/socket/unix/transport/connectioned.go @@ -252,7 +252,7 @@ func (e *connectionedEndpoint) Close() { // BidirectionalConnect implements BoundEndpoint.BidirectionalConnect. func (e *connectionedEndpoint) BidirectionalConnect(ctx context.Context, ce ConnectingEndpoint, returnConnect func(Receiver, ConnectedEndpoint)) *syserr.Error { if ce.Type() != e.stype { - return syserr.ErrConnectionRefused + return syserr.ErrWrongProtocolForSocket } // Check if ce is e to avoid a deadlock. diff --git a/pkg/sentry/socket/unix/unix.go b/pkg/sentry/socket/unix/unix.go index 5b29e9d7f..4bb2b6ff4 100644 --- a/pkg/sentry/socket/unix/unix.go +++ b/pkg/sentry/socket/unix/unix.go @@ -417,7 +417,18 @@ func (s *socketOpsCommon) Connect(t *kernel.Task, sockaddr []byte, blocking bool defer ep.Release() // Connect the server endpoint. - return s.ep.Connect(t, ep) + err = s.ep.Connect(t, ep) + + if err == syserr.ErrWrongProtocolForSocket { + // Linux for abstract sockets returns ErrConnectionRefused + // instead of ErrWrongProtocolForSocket. + path, _ := extractPath(sockaddr) + if len(path) > 0 && path[0] == 0 { + err = syserr.ErrConnectionRefused + } + } + + return err } // Write implements fs.FileOperations.Write. @@ -448,15 +459,25 @@ func (s *socketOpsCommon) SendMsg(t *kernel.Task, src usermem.IOSequence, to []b To: nil, } if len(to) > 0 { - ep, err := extractEndpoint(t, to) - if err != nil { - return 0, err - } - defer ep.Release() - w.To = ep + switch s.stype { + case linux.SOCK_SEQPACKET: + to = nil + case linux.SOCK_STREAM: + if s.State() == linux.SS_CONNECTED { + return 0, syserr.ErrAlreadyConnected + } + return 0, syserr.ErrNotSupported + default: + ep, err := extractEndpoint(t, to) + if err != nil { + return 0, err + } + defer ep.Release() + w.To = ep - if ep.Passcred() && w.Control.Credentials == nil { - w.Control.Credentials = control.MakeCreds(t) + if ep.Passcred() && w.Control.Credentials == nil { + w.Control.Credentials = control.MakeCreds(t) + } } } diff --git a/pkg/sentry/socket/unix/unix_vfs2.go b/pkg/sentry/socket/unix/unix_vfs2.go index 45e109361..8c32371a2 100644 --- a/pkg/sentry/socket/unix/unix_vfs2.go +++ b/pkg/sentry/socket/unix/unix_vfs2.go @@ -26,6 +26,7 @@ import ( "gvisor.dev/gvisor/pkg/sentry/socket/netstack" "gvisor.dev/gvisor/pkg/sentry/socket/unix/transport" "gvisor.dev/gvisor/pkg/sentry/vfs" + "gvisor.dev/gvisor/pkg/sentry/vfs/lock" "gvisor.dev/gvisor/pkg/syserr" "gvisor.dev/gvisor/pkg/syserror" "gvisor.dev/gvisor/pkg/tcpip" @@ -39,6 +40,7 @@ type SocketVFS2 struct { vfsfd vfs.FileDescription vfs.FileDescriptionDefaultImpl vfs.DentryMetadataFileDescriptionImpl + vfs.LockFD socketOpsCommon } @@ -51,7 +53,7 @@ func NewSockfsFile(t *kernel.Task, ep transport.Endpoint, stype linux.SockType) mnt := t.Kernel().SocketMount() d := sockfs.NewDentry(t.Credentials(), mnt) - fd, err := NewFileDescription(ep, stype, linux.O_RDWR, mnt, d) + fd, err := NewFileDescription(ep, stype, linux.O_RDWR, mnt, d, &lock.FileLocks{}) if err != nil { return nil, syserr.FromError(err) } @@ -60,7 +62,7 @@ func NewSockfsFile(t *kernel.Task, ep transport.Endpoint, stype linux.SockType) // NewFileDescription creates and returns a socket file description // corresponding to the given mount and dentry. -func NewFileDescription(ep transport.Endpoint, stype linux.SockType, flags uint32, mnt *vfs.Mount, d *vfs.Dentry) (*vfs.FileDescription, error) { +func NewFileDescription(ep transport.Endpoint, stype linux.SockType, flags uint32, mnt *vfs.Mount, d *vfs.Dentry, locks *lock.FileLocks) (*vfs.FileDescription, error) { // You can create AF_UNIX, SOCK_RAW sockets. They're the same as // SOCK_DGRAM and don't require CAP_NET_RAW. if stype == linux.SOCK_RAW { @@ -73,6 +75,7 @@ func NewFileDescription(ep transport.Endpoint, stype linux.SockType, flags uint3 stype: stype, }, } + sock.LockFD.Init(locks) vfsfd := &sock.vfsfd if err := vfsfd.Init(sock, flags, mnt, d, &vfs.FileDescriptionOptions{ DenyPRead: true, |