Compare commits

..

2 Commits

Author SHA1 Message Date
Jan Schär d1b403c81b Deprecate Rule.Flags field
The functionality added in a46119e5 never worked: If you set
NFTA_RULE_POSITION to 0, the kernel will just complain that a rule with
this handle does not exist. This removes the broken functionality,
leaving the field deprecated.

https://github.com/google/nftables/issues/126 mentions that the nft
command allows referring to rules by index. But here is a quote from the
nft manpage:

> The add and insert commands support an optional location specifier,
> which is either a handle or the index (starting at zero) of an
> existing rule. Internally, rule locations are always identified by
> handle and the translation from index happens in userspace.

In other words, identifiying rules by index is a feature of nft and is
not part of the kernel interface.
2025-02-20 20:26:56 +01:00
Jan Schär 6db9cddf45 Set rule handle during flush
This change makes it possible to delete rules after inserting them,
without needing to query the rules first. Rules can be deleted both
before and after they are flushed. Additionally, this allows positioning
a new rule next to an existing rule, both before and after the existing
rule is flushed.

There are two ways to refer to a rule: Either by ID or by handle. The ID
is assigned by userspace, and is only valid within a transaction, so it
can only be used before the flush. The handle is assigned by the kernel
when the transaction is committed, and can thus only be used after the
flush. We thus need to set an ID on each newly created rule, and
retrieve the handle of the rule during the flush.

There was an existing mechanism to allocate IDs for sets, but this was
using a global counter without any synchronization to prevent data
races. I replaced this by a new mechanism which uses a connection-scoped
counter.

I implemented a new mechanism for retrieving replies in Flush, and
handling these replies by adding a callback to netlink messages. There
was some existing code to handle "overrun", which I deleted, because it
was nonsensical and just worked by accident. NLMSG_OVERRUN is in fact
not a flag, but a complete message type, so the (re&netlink.Overrun)
masking makes no sense. Even better, NLMSG_OVERRUN is never actually
used by Linux. What this code was actually doing was skipping over the
NFT_MSG_NEWRULE replies, and possibly a NFT_MSG_NEWGEN reply.

I had to update all existing tests which compared generated netlink
messages against a reference, by inserting the newly added ID attribute.
We also need to generate replies for the NFT_MSG_NEWRULE messages with a
handle added.
2025-02-20 20:26:56 +01:00
8 changed files with 83 additions and 146 deletions

View File

@ -186,9 +186,6 @@ func (e *Ct) unmarshal(fam byte, data []byte) error {
e.Register = ad.Uint32()
case unix.NFTA_CT_DIRECTION:
e.Direction = ad.Uint32()
case unix.NFTA_CT_SREG:
e.SourceRegister = true
e.Register = ad.Uint32()
}
}
return ad.Err()

View File

@ -38,14 +38,6 @@ func TestCt(t *testing.T) {
Direction: 1, // direction: reply
},
},
{
name: "Unmarshal Ct source register case",
ct: Ct{
Register: 1,
Key: CtKeySRC,
SourceRegister: true,
},
},
}
for _, tt := range tests {

3
gen.go
View File

@ -3,7 +3,6 @@ package nftables
import (
"encoding/binary"
"fmt"
"github.com/mdlayher/netlink"
"golang.org/x/sys/unix"
)
@ -14,7 +13,7 @@ type GenMsg struct {
ProcComm string // [16]byte - max 16bytes - kernel TASK_COMM_LEN
}
const genHeaderType = netlink.HeaderType((unix.NFNL_SUBSYS_NFTABLES << 8) | unix.NFT_MSG_NEWGEN)
var genHeaderType = netlink.HeaderType((unix.NFNL_SUBSYS_NFTABLES << 8) | unix.NFT_MSG_NEWGEN)
func genFromMsg(msg netlink.Message) (*GenMsg, error) {
if got, want := msg.Header.Type, genHeaderType; got != want {

View File

@ -733,14 +733,6 @@ func TestMasqMarshalUnmarshal(t *testing.T) {
Table: filter,
Chain: postrouting,
Exprs: []expr.Any{
&expr.Immediate{
Register: min,
Data: binaryutil.BigEndian.PutUint16(4070),
},
&expr.Immediate{
Register: max,
Data: binaryutil.BigEndian.PutUint16(4090),
},
&expr.Masq{
ToPorts: true,
RegProtoMin: min,
@ -771,13 +763,13 @@ func TestMasqMarshalUnmarshal(t *testing.T) {
}
rule := rules[0]
if got, want := len(rule.Exprs), 3; got != want {
if got, want := len(rule.Exprs), 1; got != want {
t.Fatalf("unexpected number of exprs: got %d, want %d", got, want)
}
me, ok := rule.Exprs[2].(*expr.Masq)
me, ok := rule.Exprs[0].(*expr.Masq)
if !ok {
t.Fatalf("unexpected expression type: got %T, want *expr.Masq", rule.Exprs[2])
t.Fatalf("unexpected expression type: got %T, want *expr.Masq", rule.Exprs[0])
}
if got, want := me.ToPorts, true; got != want {
@ -3619,58 +3611,7 @@ func TestIP6SetAddElements(t *testing.T) {
t.Errorf("c.GetSetElements(portSet) failed: %v", err)
}
if len(elements) != 2 {
t.Fatalf("len(portSetElements) = %d, want 2", len(elements))
}
}
func TestSetElementBatching(t *testing.T) {
// Create a new network namespace to test these operations,
// and tear down the namespace at test completion.
c, newNS := nftest.OpenSystemConn(t, *enableSysTests)
defer nftest.CleanupSystemConn(t, newNS)
// Clear all rules at the beginning + end of the test.
c.FlushRuleset()
defer c.FlushRuleset()
filter := c.AddTable(&nftables.Table{
Family: nftables.TableFamilyIPv4,
Name: "filter",
})
portSet := &nftables.Set{
Table: filter,
Name: "ports",
KeyType: nftables.TypeInetService,
}
// The 5000 elements will need to be split into 3 batches to make each batch
// fit into a message.
elements := make([]nftables.SetElement, 5000)
for i := range elements {
elements[i].Key = binaryutil.BigEndian.PutUint16(uint16(i))
elements[i].Comment = "0123456789"
}
if err := c.AddSet(portSet, elements); err != nil {
t.Errorf("c.AddSet(portSet) failed: %v", err)
}
if err := c.Flush(); err != nil {
t.Errorf("c.Flush() failed: %v", err)
}
gotElements, err := c.GetSetElements(portSet)
if err != nil {
t.Errorf("c.GetSetElements(portSet) failed: %v", err)
}
if len(gotElements) != len(elements) {
t.Errorf("len(gotElements) = %d, want %d", len(gotElements), len(elements))
}
gotNumbers := make([]bool, len(elements))
for _, element := range gotElements {
gotNumbers[binaryutil.BigEndian.Uint16(element.Key)] = true
}
for i := range gotNumbers {
if !gotNumbers[i] {
t.Errorf("Missing element %d", i)
break
}
t.Fatalf("len(portSetElements) = %d, want 2", len(sets))
}
}

2
obj.go
View File

@ -25,7 +25,7 @@ import (
"golang.org/x/sys/unix"
)
const (
var (
newObjHeaderType = netlink.HeaderType((unix.NFNL_SUBSYS_NFTABLES << 8) | unix.NFT_MSG_NEWOBJ)
delObjHeaderType = netlink.HeaderType((unix.NFNL_SUBSYS_NFTABLES << 8) | unix.NFT_MSG_DELOBJ)
)

View File

@ -25,7 +25,7 @@ import (
"golang.org/x/sys/unix"
)
const (
var (
newRuleHeaderType = netlink.HeaderType((unix.NFNL_SUBSYS_NFTABLES << 8) | unix.NFT_MSG_NEWRULE)
delRuleHeaderType = netlink.HeaderType((unix.NFNL_SUBSYS_NFTABLES << 8) | unix.NFT_MSG_DELRULE)
)

132
set.go
View File

@ -18,7 +18,6 @@ import (
"encoding/binary"
"errors"
"fmt"
"math"
"strings"
"time"
@ -165,9 +164,7 @@ var (
TypeTimeDay,
TypeCGroupV2,
}
)
const (
// ctLabelBitSize is defined in https://git.netfilter.org/nftables/tree/src/ct.c.
ctLabelBitSize uint32 = 128
@ -378,31 +375,24 @@ func (cc *Conn) SetAddElements(s *Set, vals []SetElement) error {
if s.Anonymous {
return errors.New("anonymous sets cannot be updated")
}
return cc.appendElemList(s, vals, unix.NFT_MSG_NEWSETELEM)
}
// SetDeleteElements deletes data points from an nftables set.
func (cc *Conn) SetDeleteElements(s *Set, vals []SetElement) error {
cc.mu.Lock()
defer cc.mu.Unlock()
if s.Anonymous {
return errors.New("anonymous sets cannot be updated")
elements, err := s.makeElemList(vals, s.ID)
if err != nil {
return err
}
return cc.appendElemList(s, vals, unix.NFT_MSG_DELSETELEM)
}
cc.messages = append(cc.messages, netlinkMessage{
Header: netlink.Header{
Type: netlink.HeaderType((unix.NFNL_SUBSYS_NFTABLES << 8) | unix.NFT_MSG_NEWSETELEM),
Flags: netlink.Request | netlink.Acknowledge | netlink.Create,
},
Data: append(extraHeader(uint8(s.Table.Family), 0), cc.marshalAttr(elements)...),
})
// maxElemBatchSize is the maximum size in bytes of encoded set elements which
// are sent in one netlink message. The size field of a netlink attribute is a
// uint16, and 1024 bytes is more than enough for the per-message headers.
const maxElemBatchSize = math.MaxUint16 - 1024
func (cc *Conn) appendElemList(s *Set, vals []SetElement, hdrType uint16) error {
if len(vals) == 0 {
return nil
}
}
func (s *Set) makeElemList(vals []SetElement, id uint32) ([]netlink.Attribute, error) {
var elements []netlink.Attribute
batchSize := 0
var batches [][]netlink.Attribute
for i, v := range vals {
item := make([]netlink.Attribute, 0)
@ -414,14 +404,14 @@ func (cc *Conn) appendElemList(s *Set, vals []SetElement, hdrType uint16) error
encodedKey, err := netlink.MarshalAttributes([]netlink.Attribute{{Type: unix.NFTA_DATA_VALUE, Data: v.Key}})
if err != nil {
return fmt.Errorf("marshal key %d: %v", i, err)
return nil, fmt.Errorf("marshal key %d: %v", i, err)
}
item = append(item, netlink.Attribute{Type: unix.NFTA_SET_ELEM_KEY | unix.NLA_F_NESTED, Data: encodedKey})
if len(v.KeyEnd) > 0 {
encodedKeyEnd, err := netlink.MarshalAttributes([]netlink.Attribute{{Type: unix.NFTA_DATA_VALUE, Data: v.KeyEnd}})
if err != nil {
return fmt.Errorf("marshal key end %d: %v", i, err)
return nil, fmt.Errorf("marshal key end %d: %v", i, err)
}
item = append(item, netlink.Attribute{Type: NFTA_SET_ELEM_KEY_END | unix.NLA_F_NESTED, Data: encodedKeyEnd})
}
@ -441,7 +431,7 @@ func (cc *Conn) appendElemList(s *Set, vals []SetElement, hdrType uint16) error
{Type: unix.NFTA_DATA_VALUE, Data: binaryutil.BigEndian.PutUint32(uint32(v.VerdictData.Kind))},
})
if err != nil {
return fmt.Errorf("marshal item %d: %v", i, err)
return nil, fmt.Errorf("marshal item %d: %v", i, err)
}
encodedVal = append(encodedVal, encodedKind...)
if len(v.VerdictData.Chain) != 0 {
@ -449,21 +439,21 @@ func (cc *Conn) appendElemList(s *Set, vals []SetElement, hdrType uint16) error
{Type: unix.NFTA_SET_ELEM_DATA, Data: []byte(v.VerdictData.Chain + "\x00")},
})
if err != nil {
return fmt.Errorf("marshal item %d: %v", i, err)
return nil, fmt.Errorf("marshal item %d: %v", i, err)
}
encodedVal = append(encodedVal, encodedChain...)
}
encodedVerdict, err := netlink.MarshalAttributes([]netlink.Attribute{
{Type: unix.NFTA_SET_ELEM_DATA | unix.NLA_F_NESTED, Data: encodedVal}})
if err != nil {
return fmt.Errorf("marshal item %d: %v", i, err)
return nil, fmt.Errorf("marshal item %d: %v", i, err)
}
item = append(item, netlink.Attribute{Type: unix.NFTA_SET_ELEM_DATA | unix.NLA_F_NESTED, Data: encodedVerdict})
case len(v.Val) > 0:
// Since v.Val's length is not 0 then, v is a regular map element, need to add to the attributes
encodedVal, err := netlink.MarshalAttributes([]netlink.Attribute{{Type: unix.NFTA_DATA_VALUE, Data: v.Val}})
if err != nil {
return fmt.Errorf("marshal item %d: %v", i, err)
return nil, fmt.Errorf("marshal item %d: %v", i, err)
}
item = append(item, netlink.Attribute{Type: unix.NFTA_SET_ELEM_DATA | unix.NLA_F_NESTED, Data: encodedVal})
@ -479,42 +469,22 @@ func (cc *Conn) appendElemList(s *Set, vals []SetElement, hdrType uint16) error
encodedItem, err := netlink.MarshalAttributes(item)
if err != nil {
return fmt.Errorf("marshal item %d: %v", i, err)
}
itemSize := unix.NLA_HDRLEN + len(encodedItem)
if batchSize+itemSize > maxElemBatchSize {
batches = append(batches, elements)
elements = nil
batchSize = 0
return nil, fmt.Errorf("marshal item %d: %v", i, err)
}
elements = append(elements, netlink.Attribute{Type: uint16(i+1) | unix.NLA_F_NESTED, Data: encodedItem})
batchSize += itemSize
}
batches = append(batches, elements)
for _, batch := range batches {
encodedElem, err := netlink.MarshalAttributes(batch)
encodedElem, err := netlink.MarshalAttributes(elements)
if err != nil {
return fmt.Errorf("marshal elements: %v", err)
return nil, fmt.Errorf("marshal elements: %v", err)
}
message := []netlink.Attribute{
{Type: unix.NFTA_SET_ELEM_LIST_SET, Data: []byte(s.Name + "\x00")},
{Type: unix.NFTA_SET_ELEM_LIST_SET_ID, Data: binaryutil.BigEndian.PutUint32(s.ID)},
{Type: unix.NFTA_SET_ELEM_LIST_TABLE, Data: []byte(s.Table.Name + "\x00")},
return []netlink.Attribute{
{Type: unix.NFTA_SET_NAME, Data: []byte(s.Name + "\x00")},
{Type: unix.NFTA_LOOKUP_SET_ID, Data: binaryutil.BigEndian.PutUint32(id)},
{Type: unix.NFTA_SET_TABLE, Data: []byte(s.Table.Name + "\x00")},
{Type: unix.NFTA_SET_ELEM_LIST_ELEMENTS | unix.NLA_F_NESTED, Data: encodedElem},
}
cc.messages = append(cc.messages, netlinkMessage{
Header: netlink.Header{
Type: netlink.HeaderType((unix.NFNL_SUBSYS_NFTABLES << 8) | hdrType),
Flags: netlink.Request | netlink.Acknowledge | netlink.Create,
},
Data: append(extraHeader(uint8(s.Table.Family), 0), cc.marshalAttr(message)...),
})
}
return nil
}, nil
}
// AddSet adds the specified Set.
@ -689,7 +659,22 @@ func (cc *Conn) AddSet(s *Set, vals []SetElement) error {
})
// Set the values of the set if initial values were provided.
return cc.appendElemList(s, vals, unix.NFT_MSG_NEWSETELEM)
if len(vals) > 0 {
hdrType := unix.NFT_MSG_NEWSETELEM
elements, err := s.makeElemList(vals, s.ID)
if err != nil {
return err
}
cc.messages = append(cc.messages, netlinkMessage{
Header: netlink.Header{
Type: netlink.HeaderType((unix.NFNL_SUBSYS_NFTABLES << 8) | hdrType),
Flags: netlink.Request | netlink.Acknowledge | netlink.Create,
},
Data: append(extraHeader(uint8(s.Table.Family), 0), cc.marshalAttr(elements)...),
})
}
return nil
}
// DelSet deletes a specific set, along with all elements it contains.
@ -709,6 +694,29 @@ func (cc *Conn) DelSet(s *Set) {
})
}
// SetDeleteElements deletes data points from an nftables set.
func (cc *Conn) SetDeleteElements(s *Set, vals []SetElement) error {
cc.mu.Lock()
defer cc.mu.Unlock()
if s.Anonymous {
return errors.New("anonymous sets cannot be updated")
}
elements, err := s.makeElemList(vals, s.ID)
if err != nil {
return err
}
cc.messages = append(cc.messages, netlinkMessage{
Header: netlink.Header{
Type: netlink.HeaderType((unix.NFNL_SUBSYS_NFTABLES << 8) | unix.NFT_MSG_DELSETELEM),
Flags: netlink.Request | netlink.Acknowledge | netlink.Create,
},
Data: append(extraHeader(uint8(s.Table.Family), 0), cc.marshalAttr(elements)...),
})
return nil
}
// FlushSet deletes all data points from an nftables set.
func (cc *Conn) FlushSet(s *Set) {
cc.mu.Lock()
@ -726,7 +734,7 @@ func (cc *Conn) FlushSet(s *Set) {
})
}
const (
var (
newSetHeaderType = netlink.HeaderType((unix.NFNL_SUBSYS_NFTABLES << 8) | unix.NFT_MSG_NEWSET)
delSetHeaderType = netlink.HeaderType((unix.NFNL_SUBSYS_NFTABLES << 8) | unix.NFT_MSG_DELSET)
)
@ -826,7 +834,7 @@ func parseSetDatatype(magic uint32) (SetDatatype, error) {
return dt, nil
}
const (
var (
newElemHeaderType = netlink.HeaderType((unix.NFNL_SUBSYS_NFTABLES << 8) | unix.NFT_MSG_NEWSETELEM)
delElemHeaderType = netlink.HeaderType((unix.NFNL_SUBSYS_NFTABLES << 8) | unix.NFT_MSG_DELSETELEM)
)
@ -964,8 +972,8 @@ func (cc *Conn) GetSetElements(s *Set) ([]SetElement, error) {
defer func() { _ = closer() }()
data, err := netlink.MarshalAttributes([]netlink.Attribute{
{Type: unix.NFTA_SET_ELEM_LIST_TABLE, Data: []byte(s.Table.Name + "\x00")},
{Type: unix.NFTA_SET_ELEM_LIST_SET, Data: []byte(s.Name + "\x00")},
{Type: unix.NFTA_SET_TABLE, Data: []byte(s.Table.Name + "\x00")},
{Type: unix.NFTA_SET_NAME, Data: []byte(s.Name + "\x00")},
})
if err != nil {
return nil, err

View File

@ -21,7 +21,7 @@ import (
"golang.org/x/sys/unix"
)
const (
var (
newTableHeaderType = netlink.HeaderType((unix.NFNL_SUBSYS_NFTABLES << 8) | unix.NFT_MSG_NEWTABLE)
delTableHeaderType = netlink.HeaderType((unix.NFNL_SUBSYS_NFTABLES << 8) | unix.NFT_MSG_DELTABLE)
)