Merge release-20200810.0-65-g182f66ee5 (automated)

9 files changed, 306 insertions, 247 deletions

diff --git a/pkg/abi/linux/linux_abi_autogen_unsafe.go b/pkg/abi/linux/linux_abi_autogen_unsafe.go
index 7f3866f36..133e435c6 100644
--- a/pkg/abi/linux/linux_abi_autogen_unsafe.go
+++ b/pkg/abi/linux/linux_abi_autogen_unsafe.go
@@ -627,7 +627,7 @@ func (f *FUSEHeaderIn) UnmarshalBytes(src []byte) {
 // Packed implements marshal.Marshallable.Packed.
 //go:nosplit
 func (f *FUSEHeaderIn) Packed() bool {
-    return f.Unique.Packed() && f.Opcode.Packed()
+    return f.Opcode.Packed() && f.Unique.Packed()
 }
 
 // MarshalUnsafe implements marshal.Marshallable.MarshalUnsafe.
@@ -683,7 +683,7 @@ func (f *FUSEHeaderIn) CopyOut(task marshal.Task, addr usermem.Addr) (int, error
 // CopyIn implements marshal.Marshallable.CopyIn.
 //go:nosplit
 func (f *FUSEHeaderIn) CopyIn(task marshal.Task, addr usermem.Addr) (int, error) {
-    if !f.Unique.Packed() && f.Opcode.Packed() {
+    if !f.Opcode.Packed() && f.Unique.Packed() {
         // Type FUSEHeaderIn doesn't have a packed layout in memory, fall back to UnmarshalBytes.
         buf := task.CopyScratchBuffer(f.SizeBytes()) // escapes: okay.
         length, err := task.CopyInBytes(addr, buf) // escapes: okay.
@@ -2223,7 +2223,7 @@ func (i *IPTIP) MarshalUnsafe(dst []byte) {
 
 // UnmarshalUnsafe implements marshal.Marshallable.UnmarshalUnsafe.
 func (i *IPTIP) UnmarshalUnsafe(src []byte) {
-    if i.Dst.Packed() && i.SrcMask.Packed() && i.DstMask.Packed() && i.Src.Packed() {
+    if i.DstMask.Packed() && i.Src.Packed() && i.Dst.Packed() && i.SrcMask.Packed() {
         safecopy.CopyOut(unsafe.Pointer(i), src)
     } else {
         // Type IPTIP doesn't have a packed layout in memory, fallback to UnmarshalBytes.
@@ -2290,7 +2290,7 @@ func (i *IPTIP) CopyIn(task marshal.Task, addr usermem.Addr) (int, error) {
 
 // WriteTo implements io.WriterTo.WriteTo.
 func (i *IPTIP) WriteTo(w io.Writer) (int64, error) {
-    if !i.DstMask.Packed() && i.Src.Packed() && i.Dst.Packed() && i.SrcMask.Packed() {
+    if !i.Src.Packed() && i.Dst.Packed() && i.SrcMask.Packed() && i.DstMask.Packed() {
         // Type IPTIP doesn't have a packed layout in memory, fall back to MarshalBytes.
         buf := make([]byte, i.SizeBytes())
         i.MarshalBytes(buf)
@@ -2999,7 +2999,7 @@ func (i *IP6TEntry) UnmarshalBytes(src []byte) {
 // Packed implements marshal.Marshallable.Packed.
 //go:nosplit
 func (i *IP6TEntry) Packed() bool {
-    return i.Counters.Packed() && i.IPv6.Packed()
+    return i.IPv6.Packed() && i.Counters.Packed()
 }
 
 // MarshalUnsafe implements marshal.Marshallable.MarshalUnsafe.
@@ -3196,7 +3196,7 @@ func (i *IP6TIP) UnmarshalBytes(src []byte) {
 // Packed implements marshal.Marshallable.Packed.
 //go:nosplit
 func (i *IP6TIP) Packed() bool {
-    return i.Src.Packed() && i.Dst.Packed() && i.SrcMask.Packed() && i.DstMask.Packed()
+    return i.DstMask.Packed() && i.Src.Packed() && i.Dst.Packed() && i.SrcMask.Packed()
 }
 
 // MarshalUnsafe implements marshal.Marshallable.MarshalUnsafe.
@@ -3211,7 +3211,7 @@ func (i *IP6TIP) MarshalUnsafe(dst []byte) {
 
 // UnmarshalUnsafe implements marshal.Marshallable.UnmarshalUnsafe.
 func (i *IP6TIP) UnmarshalUnsafe(src []byte) {
-    if i.SrcMask.Packed() && i.DstMask.Packed() && i.Src.Packed() && i.Dst.Packed() {
+    if i.Dst.Packed() && i.SrcMask.Packed() && i.DstMask.Packed() && i.Src.Packed() {
         safecopy.CopyOut(unsafe.Pointer(i), src)
     } else {
         // Type IP6TIP doesn't have a packed layout in memory, fallback to UnmarshalBytes.
@@ -3222,7 +3222,7 @@ func (i *IP6TIP) UnmarshalUnsafe(src []byte) {
 // CopyOutN implements marshal.Marshallable.CopyOutN.
 //go:nosplit
 func (i *IP6TIP) CopyOutN(task marshal.Task, addr usermem.Addr, limit int) (int, error) {
-    if !i.DstMask.Packed() && i.Src.Packed() && i.Dst.Packed() && i.SrcMask.Packed() {
+    if !i.Src.Packed() && i.Dst.Packed() && i.SrcMask.Packed() && i.DstMask.Packed() {
         // Type IP6TIP doesn't have a packed layout in memory, fall back to MarshalBytes.
         buf := task.CopyScratchBuffer(i.SizeBytes()) // escapes: okay.
         i.MarshalBytes(buf) // escapes: fallback.
@@ -3252,7 +3252,7 @@ func (i *IP6TIP) CopyOut(task marshal.Task, addr usermem.Addr) (int, error) {
 // CopyIn implements marshal.Marshallable.CopyIn.
 //go:nosplit
 func (i *IP6TIP) CopyIn(task marshal.Task, addr usermem.Addr) (int, error) {
-    if !i.DstMask.Packed() && i.Src.Packed() && i.Dst.Packed() && i.SrcMask.Packed() {
+    if !i.Src.Packed() && i.Dst.Packed() && i.SrcMask.Packed() && i.DstMask.Packed() {
         // Type IP6TIP doesn't have a packed layout in memory, fall back to UnmarshalBytes.
         buf := task.CopyScratchBuffer(i.SizeBytes()) // escapes: okay.
         length, err := task.CopyInBytes(addr, buf) // escapes: okay.
diff --git a/pkg/sentry/fsimpl/tmpfs/inode_refs.go b/pkg/sentry/fsimpl/tmpfs/inode_refs.go
index 8b7ff185f..175dc3409 100644
--- a/pkg/sentry/fsimpl/tmpfs/inode_refs.go
+++ b/pkg/sentry/fsimpl/tmpfs/inode_refs.go
@@ -1,11 +1,11 @@
 package tmpfs
 
 import (
+	refs_vfs1 "gvisor.dev/gvisor/pkg/refs"
 	"runtime"
 	"sync/atomic"
 
 	"gvisor.dev/gvisor/pkg/log"
-	refs_vfs1 "gvisor.dev/gvisor/pkg/refs"
 )
 
 // ownerType is used to customize logging. Note that we use a pointer to T so
diff --git a/pkg/sentry/platform/ring0/defs_impl_arm64.go b/pkg/sentry/platform/ring0/defs_impl_arm64.go
index eda1e1484..424b66f76 100644
--- a/pkg/sentry/platform/ring0/defs_impl_arm64.go
+++ b/pkg/sentry/platform/ring0/defs_impl_arm64.go
@@ -1,13 +1,13 @@
 package ring0
 
 import (
-	"fmt"
 	"gvisor.dev/gvisor/pkg/sentry/arch"
 	"gvisor.dev/gvisor/pkg/sentry/platform/ring0/pagetables"
-	"io"
-	"reflect"
 
+	"fmt"
 	"gvisor.dev/gvisor/pkg/usermem"
+	"io"
+	"reflect"
 )
 
 // Useful bits.
diff --git a/pkg/sentry/socket/netfilter/ipv4.go b/pkg/sentry/socket/netfilter/ipv4.go
new file mode 100644
index 000000000..4fb887e49
--- /dev/null
+++ b/pkg/sentry/socket/netfilter/ipv4.go
@@ -0,0 +1,235 @@
+// Copyright 2020 The gVisor Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+package netfilter
+
+import (
+	"bytes"
+	"fmt"
+
+	"gvisor.dev/gvisor/pkg/abi/linux"
+	"gvisor.dev/gvisor/pkg/binary"
+	"gvisor.dev/gvisor/pkg/syserr"
+	"gvisor.dev/gvisor/pkg/tcpip"
+	"gvisor.dev/gvisor/pkg/tcpip/header"
+	"gvisor.dev/gvisor/pkg/tcpip/stack"
+	"gvisor.dev/gvisor/pkg/usermem"
+)
+
+// emptyIPv4Filter is for comparison with a rule's filters to determine whether
+// it is also empty. It is immutable.
+var emptyIPv4Filter = stack.IPHeaderFilter{
+	Dst:     "\x00\x00\x00\x00",
+	DstMask: "\x00\x00\x00\x00",
+	Src:     "\x00\x00\x00\x00",
+	SrcMask: "\x00\x00\x00\x00",
+}
+
+func getEntries4(table stack.Table, info *linux.IPTGetinfo) linux.KernelIPTGetEntries {
+	var entries linux.KernelIPTGetEntries
+	copy(entries.Name[:], info.Name[:])
+
+	for ruleIdx, rule := range table.Rules {
+		nflog("convert to binary: current offset: %d", entries.Size)
+
+		setHooksAndUnderflow(info, table, entries.Size, ruleIdx)
+		// Each rule corresponds to an entry.
+		entry := linux.KernelIPTEntry{
+			Entry: linux.IPTEntry{
+				IP: linux.IPTIP{
+					Protocol: uint16(rule.Filter.Protocol),
+				},
+				NextOffset:   linux.SizeOfIPTEntry,
+				TargetOffset: linux.SizeOfIPTEntry,
+			},
+		}
+		copy(entry.Entry.IP.Dst[:], rule.Filter.Dst)
+		copy(entry.Entry.IP.DstMask[:], rule.Filter.DstMask)
+		copy(entry.Entry.IP.Src[:], rule.Filter.Src)
+		copy(entry.Entry.IP.SrcMask[:], rule.Filter.SrcMask)
+		copy(entry.Entry.IP.OutputInterface[:], rule.Filter.OutputInterface)
+		copy(entry.Entry.IP.OutputInterfaceMask[:], rule.Filter.OutputInterfaceMask)
+		if rule.Filter.DstInvert {
+			entry.Entry.IP.InverseFlags |= linux.IPT_INV_DSTIP
+		}
+		if rule.Filter.SrcInvert {
+			entry.Entry.IP.InverseFlags |= linux.IPT_INV_SRCIP
+		}
+		if rule.Filter.OutputInterfaceInvert {
+			entry.Entry.IP.InverseFlags |= linux.IPT_INV_VIA_OUT
+		}
+
+		for _, matcher := range rule.Matchers {
+			// Serialize the matcher and add it to the
+			// entry.
+			serialized := marshalMatcher(matcher)
+			nflog("convert to binary: matcher serialized as: %v", serialized)
+			if len(serialized)%8 != 0 {
+				panic(fmt.Sprintf("matcher %T is not 64-bit aligned", matcher))
+			}
+			entry.Elems = append(entry.Elems, serialized...)
+			entry.Entry.NextOffset += uint16(len(serialized))
+			entry.Entry.TargetOffset += uint16(len(serialized))
+		}
+
+		// Serialize and append the target.
+		serialized := marshalTarget(rule.Target)
+		if len(serialized)%8 != 0 {
+			panic(fmt.Sprintf("target %T is not 64-bit aligned", rule.Target))
+		}
+		entry.Elems = append(entry.Elems, serialized...)
+		entry.Entry.NextOffset += uint16(len(serialized))
+
+		nflog("convert to binary: adding entry: %+v", entry)
+
+		entries.Size += uint32(entry.Entry.NextOffset)
+		entries.Entrytable = append(entries.Entrytable, entry)
+		info.NumEntries++
+	}
+
+	info.Size = entries.Size
+	nflog("convert to binary: finished with an marshalled size of %d", info.Size)
+	return entries
+}
+
+func modifyEntries4(stk *stack.Stack, optVal []byte, replace *linux.IPTReplace, table *stack.Table) (map[uint32]int, *syserr.Error) {
+	nflog("set entries: setting entries in table %q", replace.Name.String())
+
+	// Convert input into a list of rules and their offsets.
+	var offset uint32
+	// offsets maps rule byte offsets to their position in table.Rules.
+	offsets := map[uint32]int{}
+	for entryIdx := uint32(0); entryIdx < replace.NumEntries; entryIdx++ {
+		nflog("set entries: processing entry at offset %d", offset)
+
+		// Get the struct ipt_entry.
+		if len(optVal) < linux.SizeOfIPTEntry {
+			nflog("optVal has insufficient size for entry %d", len(optVal))
+			return nil, syserr.ErrInvalidArgument
+		}
+		var entry linux.IPTEntry
+		buf := optVal[:linux.SizeOfIPTEntry]
+		binary.Unmarshal(buf, usermem.ByteOrder, &entry)
+		initialOptValLen := len(optVal)
+		optVal = optVal[linux.SizeOfIPTEntry:]
+
+		if entry.TargetOffset < linux.SizeOfIPTEntry {
+			nflog("entry has too-small target offset %d", entry.TargetOffset)
+			return nil, syserr.ErrInvalidArgument
+		}
+
+		// TODO(gvisor.dev/issue/170): We should support more IPTIP
+		// filtering fields.
+		filter, err := filterFromIPTIP(entry.IP)
+		if err != nil {
+			nflog("bad iptip: %v", err)
+			return nil, syserr.ErrInvalidArgument
+		}
+
+		// TODO(gvisor.dev/issue/170): Matchers and targets can specify
+		// that they only work for certain protocols, hooks, tables.
+		// Get matchers.
+		matchersSize := entry.TargetOffset - linux.SizeOfIPTEntry
+		if len(optVal) < int(matchersSize) {
+			nflog("entry doesn't have enough room for its matchers (only %d bytes remain)", len(optVal))
+			return nil, syserr.ErrInvalidArgument
+		}
+		matchers, err := parseMatchers(filter, optVal[:matchersSize])
+		if err != nil {
+			nflog("failed to parse matchers: %v", err)
+			return nil, syserr.ErrInvalidArgument
+		}
+		optVal = optVal[matchersSize:]
+
+		// Get the target of the rule.
+		targetSize := entry.NextOffset - entry.TargetOffset
+		if len(optVal) < int(targetSize) {
+			nflog("entry doesn't have enough room for its target (only %d bytes remain)", len(optVal))
+			return nil, syserr.ErrInvalidArgument
+		}
+		target, err := parseTarget(filter, optVal[:targetSize])
+		if err != nil {
+			nflog("failed to parse target: %v", err)
+			return nil, syserr.ErrInvalidArgument
+		}
+		optVal = optVal[targetSize:]
+
+		table.Rules = append(table.Rules, stack.Rule{
+			Filter:   filter,
+			Target:   target,
+			Matchers: matchers,
+		})
+		offsets[offset] = int(entryIdx)
+		offset += uint32(entry.NextOffset)
+
+		if initialOptValLen-len(optVal) != int(entry.NextOffset) {
+			nflog("entry NextOffset is %d, but entry took up %d bytes", entry.NextOffset, initialOptValLen-len(optVal))
+			return nil, syserr.ErrInvalidArgument
+		}
+	}
+	return offsets, nil
+}
+
+func filterFromIPTIP(iptip linux.IPTIP) (stack.IPHeaderFilter, error) {
+	if containsUnsupportedFields4(iptip) {
+		return stack.IPHeaderFilter{}, fmt.Errorf("unsupported fields in struct iptip: %+v", iptip)
+	}
+	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 {
+		n = len(iptip.OutputInterface)
+	}
+	ifname := string(iptip.OutputInterface[:n])
+
+	n = bytes.IndexByte([]byte(iptip.OutputInterfaceMask[:]), 0)
+	if n == -1 {
+		n = len(iptip.OutputInterfaceMask)
+	}
+	ifnameMask := string(iptip.OutputInterfaceMask[:n])
+
+	return stack.IPHeaderFilter{
+		Protocol:              tcpip.TransportProtocolNumber(iptip.Protocol),
+		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,
+	}, nil
+}
+
+func containsUnsupportedFields4(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 emptyInterface = [linux.IFNAMSIZ]byte{}
+	// Disable any supported inverse flags.
+	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/netfilter.go b/pkg/sentry/socket/netfilter/netfilter.go
index e91b0624c..df256676f 100644
--- a/pkg/sentry/socket/netfilter/netfilter.go
+++ b/pkg/sentry/socket/netfilter/netfilter.go
@@ -17,7 +17,6 @@
 package netfilter
 
 import (
-	"bytes"
 	"errors"
 	"fmt"
 
@@ -26,8 +25,6 @@ import (
 	"gvisor.dev/gvisor/pkg/log"
 	"gvisor.dev/gvisor/pkg/sentry/kernel"
 	"gvisor.dev/gvisor/pkg/syserr"
-	"gvisor.dev/gvisor/pkg/tcpip"
-	"gvisor.dev/gvisor/pkg/tcpip/header"
 	"gvisor.dev/gvisor/pkg/tcpip/stack"
 	"gvisor.dev/gvisor/pkg/usermem"
 )
@@ -37,15 +34,6 @@ import (
 // developing iptables, but can pollute sentry logs otherwise.
 const enableLogging = false
 
-// emptyFilter is for comparison with a rule's filters to determine whether it
-// is also empty. It is immutable.
-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.
 func nflog(format string, args ...interface{}) {
 	if enableLogging && log.IsLogging(log.Debug) {
@@ -71,9 +59,9 @@ func GetInfo(t *kernel.Task, stack *stack.Stack, outPtr usermem.Addr) (linux.IPT
 	return info, nil
 }
 
-// GetEntries returns netstack's iptables rules encoded for the iptables tool.
-func GetEntries(t *kernel.Task, stack *stack.Stack, outPtr usermem.Addr, outLen int) (linux.KernelIPTGetEntries, *syserr.Error) {
-	// Read in the struct and table name.
+// GetEntries4 returns netstack's iptables rules encoded for the iptables tool.
+func GetEntries4(t *kernel.Task, stack *stack.Stack, outPtr usermem.Addr, outLen int) (linux.KernelIPTGetEntries, *syserr.Error) {
+	// Read in the ABI struct.
 	var userEntries linux.IPTGetEntries
 	if _, err := userEntries.CopyIn(t, outPtr); err != nil {
 		nflog("couldn't copy in entries %q", userEntries.Name)
@@ -99,108 +87,48 @@ func GetEntries(t *kernel.Task, stack *stack.Stack, outPtr usermem.Addr, outLen
 // format expected by the iptables tool. Linux stores each table as a binary
 // blob that can only be traversed by parsing a bit, reading some offsets,
 // jumping to those offsets, parsing again, etc.
-func convertNetstackToBinary(stack *stack.Stack, tablename linux.TableName) (linux.KernelIPTGetEntries, linux.IPTGetinfo, error) {
-	table, ok := stack.IPTables().GetTable(tablename.String())
+func convertNetstackToBinary(stk *stack.Stack, tablename linux.TableName) (linux.KernelIPTGetEntries, linux.IPTGetinfo, error) {
+	// The table name has to fit in the struct.
+	if linux.XT_TABLE_MAXNAMELEN < len(tablename) {
+		return linux.KernelIPTGetEntries{}, linux.IPTGetinfo{}, fmt.Errorf("table name %q too long", tablename)
+	}
+
+	table, ok := stk.IPTables().GetTable(tablename.String())
 	if !ok {
 		return linux.KernelIPTGetEntries{}, linux.IPTGetinfo{}, fmt.Errorf("couldn't find table %q", tablename)
 	}
 
-	var entries linux.KernelIPTGetEntries
+	// Setup the info struct.
 	var info linux.IPTGetinfo
 	info.ValidHooks = table.ValidHooks()
-
-	// The table name has to fit in the struct.
-	if linux.XT_TABLE_MAXNAMELEN < len(tablename) {
-		return linux.KernelIPTGetEntries{}, linux.IPTGetinfo{}, fmt.Errorf("table name %q too long", tablename)
-	}
 	copy(info.Name[:], tablename[:])
-	copy(entries.Name[:], tablename[:])
-
-	for ruleIdx, rule := range table.Rules {
-		nflog("convert to binary: current offset: %d", entries.Size)
-
-		// Is this a chain entry point?
-		for hook, hookRuleIdx := range table.BuiltinChains {
-			if hookRuleIdx == ruleIdx {
-				nflog("convert to binary: found hook %d at offset %d", hook, entries.Size)
-				info.HookEntry[hook] = entries.Size
-			}
-		}
-		// Is this a chain underflow point?
-		for underflow, underflowRuleIdx := range table.Underflows {
-			if underflowRuleIdx == ruleIdx {
-				nflog("convert to binary: found underflow %d at offset %d", underflow, entries.Size)
-				info.Underflow[underflow] = entries.Size
-			}
-		}
 
-		// Each rule corresponds to an entry.
-		entry := linux.KernelIPTEntry{
-			Entry: linux.IPTEntry{
-				IP: linux.IPTIP{
-					Protocol: uint16(rule.Filter.Protocol),
-				},
-				NextOffset:   linux.SizeOfIPTEntry,
-				TargetOffset: linux.SizeOfIPTEntry,
-			},
-		}
-		copy(entry.Entry.IP.Dst[:], rule.Filter.Dst)
-		copy(entry.Entry.IP.DstMask[:], rule.Filter.DstMask)
-		copy(entry.Entry.IP.Src[:], rule.Filter.Src)
-		copy(entry.Entry.IP.SrcMask[:], rule.Filter.SrcMask)
-		copy(entry.Entry.IP.OutputInterface[:], rule.Filter.OutputInterface)
-		copy(entry.Entry.IP.OutputInterfaceMask[:], rule.Filter.OutputInterfaceMask)
-		if rule.Filter.DstInvert {
-			entry.Entry.IP.InverseFlags |= linux.IPT_INV_DSTIP
-		}
-		if rule.Filter.SrcInvert {
-			entry.Entry.IP.InverseFlags |= linux.IPT_INV_SRCIP
-		}
-		if rule.Filter.OutputInterfaceInvert {
-			entry.Entry.IP.InverseFlags |= linux.IPT_INV_VIA_OUT
-		}
+	entries := getEntries4(table, &info)
+	return entries, info, nil
+}
 
-		for _, matcher := range rule.Matchers {
-			// Serialize the matcher and add it to the
-			// entry.
-			serialized := marshalMatcher(matcher)
-			nflog("convert to binary: matcher serialized as: %v", serialized)
-			if len(serialized)%8 != 0 {
-				panic(fmt.Sprintf("matcher %T is not 64-bit aligned", matcher))
-			}
-			entry.Elems = append(entry.Elems, serialized...)
-			entry.Entry.NextOffset += uint16(len(serialized))
-			entry.Entry.TargetOffset += uint16(len(serialized))
+// setHooksAndUnderflow checks whether the rule at ruleIdx is a hook entrypoint
+// or underflow, in which case it fills in info.HookEntry and info.Underflows.
+func setHooksAndUnderflow(info *linux.IPTGetinfo, table stack.Table, offset uint32, ruleIdx int) {
+	// Is this a chain entry point?
+	for hook, hookRuleIdx := range table.BuiltinChains {
+		if hookRuleIdx == ruleIdx {
+			nflog("convert to binary: found hook %d at offset %d", hook, offset)
+			info.HookEntry[hook] = offset
 		}
-
-		// Serialize and append the target.
-		serialized := marshalTarget(rule.Target)
-		if len(serialized)%8 != 0 {
-			panic(fmt.Sprintf("target %T is not 64-bit aligned", rule.Target))
+	}
+	// Is this a chain underflow point?
+	for underflow, underflowRuleIdx := range table.Underflows {
+		if underflowRuleIdx == ruleIdx {
+			nflog("convert to binary: found underflow %d at offset %d", underflow, offset)
+			info.Underflow[underflow] = offset
 		}
-		entry.Elems = append(entry.Elems, serialized...)
-		entry.Entry.NextOffset += uint16(len(serialized))
-
-		nflog("convert to binary: adding entry: %+v", entry)
-
-		entries.Size += uint32(entry.Entry.NextOffset)
-		entries.Entrytable = append(entries.Entrytable, entry)
-		info.NumEntries++
 	}
-
-	nflog("convert to binary: finished with an marshalled size of %d", info.Size)
-	info.Size = entries.Size
-	return entries, info, nil
 }
 
 // SetEntries sets iptables rules for a single table. See
 // net/ipv4/netfilter/ip_tables.c:translate_table for reference.
 func SetEntries(stk *stack.Stack, optVal []byte) *syserr.Error {
-	// Get the basic rules data (struct ipt_replace).
-	if len(optVal) < linux.SizeOfIPTReplace {
-		nflog("optVal has insufficient size for replace %d", len(optVal))
-		return syserr.ErrInvalidArgument
-	}
 	var replace linux.IPTReplace
 	replaceBuf := optVal[:linux.SizeOfIPTReplace]
 	optVal = optVal[linux.SizeOfIPTReplace:]
@@ -218,79 +146,9 @@ func SetEntries(stk *stack.Stack, optVal []byte) *syserr.Error {
 		return syserr.ErrInvalidArgument
 	}
 
-	nflog("set entries: setting entries in table %q", replace.Name.String())
-
-	// Convert input into a list of rules and their offsets.
-	var offset uint32
-	// offsets maps rule byte offsets to their position in table.Rules.
-	offsets := map[uint32]int{}
-	for entryIdx := uint32(0); entryIdx < replace.NumEntries; entryIdx++ {
-		nflog("set entries: processing entry at offset %d", offset)
-
-		// Get the struct ipt_entry.
-		if len(optVal) < linux.SizeOfIPTEntry {
-			nflog("optVal has insufficient size for entry %d", len(optVal))
-			return syserr.ErrInvalidArgument
-		}
-		var entry linux.IPTEntry
-		buf := optVal[:linux.SizeOfIPTEntry]
-		binary.Unmarshal(buf, usermem.ByteOrder, &entry)
-		initialOptValLen := len(optVal)
-		optVal = optVal[linux.SizeOfIPTEntry:]
-
-		if entry.TargetOffset < linux.SizeOfIPTEntry {
-			nflog("entry has too-small target offset %d", entry.TargetOffset)
-			return syserr.ErrInvalidArgument
-		}
-
-		// TODO(gvisor.dev/issue/170): We should support more IPTIP
-		// filtering fields.
-		filter, err := filterFromIPTIP(entry.IP)
-		if err != nil {
-			nflog("bad iptip: %v", err)
-			return syserr.ErrInvalidArgument
-		}
-
-		// TODO(gvisor.dev/issue/170): Matchers and targets can specify
-		// that they only work for certain protocols, hooks, tables.
-		// Get matchers.
-		matchersSize := entry.TargetOffset - linux.SizeOfIPTEntry
-		if len(optVal) < int(matchersSize) {
-			nflog("entry doesn't have enough room for its matchers (only %d bytes remain)", len(optVal))
-			return syserr.ErrInvalidArgument
-		}
-		matchers, err := parseMatchers(filter, optVal[:matchersSize])
-		if err != nil {
-			nflog("failed to parse matchers: %v", err)
-			return syserr.ErrInvalidArgument
-		}
-		optVal = optVal[matchersSize:]
-
-		// Get the target of the rule.
-		targetSize := entry.NextOffset - entry.TargetOffset
-		if len(optVal) < int(targetSize) {
-			nflog("entry doesn't have enough room for its target (only %d bytes remain)", len(optVal))
-			return syserr.ErrInvalidArgument
-		}
-		target, err := parseTarget(filter, optVal[:targetSize])
-		if err != nil {
-			nflog("failed to parse target: %v", err)
-			return syserr.ErrInvalidArgument
-		}
-		optVal = optVal[targetSize:]
-
-		table.Rules = append(table.Rules, stack.Rule{
-			Filter:   filter,
-			Target:   target,
-			Matchers: matchers,
-		})
-		offsets[offset] = int(entryIdx)
-		offset += uint32(entry.NextOffset)
-
-		if initialOptValLen-len(optVal) != int(entry.NextOffset) {
-			nflog("entry NextOffset is %d, but entry took up %d bytes", entry.NextOffset, initialOptValLen-len(optVal))
-			return syserr.ErrInvalidArgument
-		}
+	offsets, err := modifyEntries4(stk, optVal, &replace, &table)
+	if err != nil {
+		return err
 	}
 
 	// Go through the list of supported hooks for this table and, for each
@@ -323,7 +181,7 @@ func SetEntries(stk *stack.Stack, optVal []byte) *syserr.Error {
 		}
 	}
 
-	// Add the user chains.
+	// Check the user chains.
 	for ruleIdx, rule := range table.Rules {
 		if _, ok := rule.Target.(stack.UserChainTarget); !ok {
 			continue
@@ -404,7 +262,6 @@ func parseMatchers(filter stack.IPHeaderFilter, optVal []byte) ([]stack.Matcher,
 
 		// Check some invariants.
 		if match.MatchSize < linux.SizeOfXTEntryMatch {
-
 			return nil, fmt.Errorf("match size is too small, must be at least %d", linux.SizeOfXTEntryMatch)
 		}
 		if len(optVal) < int(match.MatchSize) {
@@ -429,64 +286,11 @@ func parseMatchers(filter stack.IPHeaderFilter, optVal []byte) ([]stack.Matcher,
 	return matchers, nil
 }
 
-func filterFromIPTIP(iptip linux.IPTIP) (stack.IPHeaderFilter, error) {
-	if containsUnsupportedFields(iptip) {
-		return stack.IPHeaderFilter{}, fmt.Errorf("unsupported fields in struct iptip: %+v", iptip)
-	}
-	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 {
-		n = len(iptip.OutputInterface)
-	}
-	ifname := string(iptip.OutputInterface[:n])
-
-	n = bytes.IndexByte([]byte(iptip.OutputInterfaceMask[:]), 0)
-	if n == -1 {
-		n = len(iptip.OutputInterfaceMask)
-	}
-	ifnameMask := string(iptip.OutputInterfaceMask[:n])
-
-	return stack.IPHeaderFilter{
-		Protocol:              tcpip.TransportProtocolNumber(iptip.Protocol),
-		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,
-	}, nil
-}
-
-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 emptyInterface = [linux.IFNAMSIZ]byte{}
-	// Disable any supported inverse flags.
-	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
-}
-
 func validUnderflow(rule stack.Rule) bool {
 	if len(rule.Matchers) != 0 {
 		return false
 	}
-	if rule.Filter != emptyFilter {
+	if rule.Filter != emptyIPv4Filter {
 		return false
 	}
 	switch rule.Target.(type) {
diff --git a/pkg/sentry/socket/netstack/netstack.go b/pkg/sentry/socket/netstack/netstack.go
index e4846bc0b..0e5913b60 100644
--- a/pkg/sentry/socket/netstack/netstack.go
+++ b/pkg/sentry/socket/netstack/netstack.go
@@ -949,6 +949,9 @@ func (s *SocketOperations) GetSockOpt(t *kernel.Task, level, name int, outPtr us
 			if outLen < linux.SizeOfIPTGetinfo {
 				return nil, syserr.ErrInvalidArgument
 			}
+			if s.family != linux.AF_INET {
+				return nil, syserr.ErrInvalidArgument
+			}
 
 			stack := inet.StackFromContext(t)
 			if stack == nil {
@@ -964,12 +967,15 @@ func (s *SocketOperations) GetSockOpt(t *kernel.Task, level, name int, outPtr us
 			if outLen < linux.SizeOfIPTGetEntries {
 				return nil, syserr.ErrInvalidArgument
 			}
+			if s.family != linux.AF_INET {
+				return nil, syserr.ErrInvalidArgument
+			}
 
 			stack := inet.StackFromContext(t)
 			if stack == nil {
 				return nil, syserr.ErrNoDevice
 			}
-			entries, err := netfilter.GetEntries(t, stack.(*Stack).Stack, outPtr, outLen)
+			entries, err := netfilter.GetEntries4(t, stack.(*Stack).Stack, outPtr, outLen)
 			if err != nil {
 				return nil, err
 			}
@@ -1650,12 +1656,15 @@ func (s *SocketOperations) SetSockOpt(t *kernel.Task, level int, name int, optVa
 		return nil
 	}
 
-	if s.skType == linux.SOCK_RAW && level == linux.IPPROTO_IP {
+	if s.skType == linux.SOCK_RAW && level == linux.SOL_IP {
 		switch name {
 		case linux.IPT_SO_SET_REPLACE:
 			if len(optVal) < linux.SizeOfIPTReplace {
 				return syserr.ErrInvalidArgument
 			}
+			if s.family != linux.AF_INET {
+				return syserr.ErrInvalidArgument
+			}
 
 			stack := inet.StackFromContext(t)
 			if stack == nil {
diff --git a/pkg/sentry/socket/netstack/netstack_vfs2.go b/pkg/sentry/socket/netstack/netstack_vfs2.go
index 3335e7430..1db8ae491 100644
--- a/pkg/sentry/socket/netstack/netstack_vfs2.go
+++ b/pkg/sentry/socket/netstack/netstack_vfs2.go
@@ -239,6 +239,9 @@ func (s *SocketVFS2) GetSockOpt(t *kernel.Task, level, name int, outPtr usermem.
 			if outLen < linux.SizeOfIPTGetinfo {
 				return nil, syserr.ErrInvalidArgument
 			}
+			if s.family != linux.AF_INET {
+				return nil, syserr.ErrInvalidArgument
+			}
 
 			stack := inet.StackFromContext(t)
 			if stack == nil {
@@ -254,12 +257,15 @@ func (s *SocketVFS2) GetSockOpt(t *kernel.Task, level, name int, outPtr usermem.
 			if outLen < linux.SizeOfIPTGetEntries {
 				return nil, syserr.ErrInvalidArgument
 			}
+			if s.family != linux.AF_INET {
+				return nil, syserr.ErrInvalidArgument
+			}
 
 			stack := inet.StackFromContext(t)
 			if stack == nil {
 				return nil, syserr.ErrNoDevice
 			}
-			entries, err := netfilter.GetEntries(t, stack.(*Stack).Stack, outPtr, outLen)
+			entries, err := netfilter.GetEntries4(t, stack.(*Stack).Stack, outPtr, outLen)
 			if err != nil {
 				return nil, err
 			}
@@ -298,12 +304,15 @@ func (s *SocketVFS2) SetSockOpt(t *kernel.Task, level int, name int, optVal []by
 		return nil
 	}
 
-	if s.skType == linux.SOCK_RAW && level == linux.IPPROTO_IP {
+	if s.skType == linux.SOCK_RAW && level == linux.SOL_IP {
 		switch name {
 		case linux.IPT_SO_SET_REPLACE:
 			if len(optVal) < linux.SizeOfIPTReplace {
 				return syserr.ErrInvalidArgument
 			}
+			if s.family != linux.AF_INET {
+				return syserr.ErrInvalidArgument
+			}
 
 			stack := inet.StackFromContext(t)
 			if stack == nil {
diff --git a/pkg/sentry/socket/unix/socket_refs.go b/pkg/sentry/socket/unix/socket_refs.go
index 4c6ec186b..090395057 100644
--- a/pkg/sentry/socket/unix/socket_refs.go
+++ b/pkg/sentry/socket/unix/socket_refs.go
@@ -1,11 +1,11 @@
 package unix
 
 import (
+	refs_vfs1 "gvisor.dev/gvisor/pkg/refs"
 	"runtime"
 	"sync/atomic"
 
 	"gvisor.dev/gvisor/pkg/log"
-	refs_vfs1 "gvisor.dev/gvisor/pkg/refs"
 )
 
 // ownerType is used to customize logging. Note that we use a pointer to T so
diff --git a/pkg/sentry/strace/socket.go b/pkg/sentry/strace/socket.go
index b51c4c941..08e97e6c4 100644
--- a/pkg/sentry/strace/socket.go
+++ b/pkg/sentry/strace/socket.go
@@ -632,6 +632,8 @@ var sockOptNames = map[uint64]abi.ValueSet{
 		linux.IPV6_UNICAST_IF:          "IPV6_UNICAST_IF",
 		linux.MCAST_MSFILTER:           "MCAST_MSFILTER",
 		linux.IPV6_ADDRFORM:            "IPV6_ADDRFORM",
+		linux.IP6T_SO_GET_INFO:         "IP6T_SO_GET_INFO",
+		linux.IP6T_SO_GET_ENTRIES:      "IP6T_SO_GET_ENTRIES",
 	},
 	linux.SOL_NETLINK: {
 		linux.NETLINK_BROADCAST_ERROR:  "NETLINK_BROADCAST_ERROR",