ebpf

func (*BpfConn) Close ¶

func (bc *BpfConn) Close() error

Close closes the socket and then detaches its rate limit, once. Later calls return the first call's result.

Close does not delete the socket's debuglet_sk_map entry. That map is an SK_STORAGE map: the entry belongs to the socket and the kernel frees it when the socket is destroyed. Until then, data the kernel still sends after close(2) (the unsent send buffer, the FIN, retransmissions) keeps this run's UUID and is dropped once Detach has removed the rate entry. Deleting the entry earlier would let that tail leave unattributed and unthrottled, and deleting it after close(2) addresses a descriptor that no longer names the socket (#412).

func (*BpfConn) LocalAddr ¶

func (bc *BpfConn) LocalAddr() net.Addr

func (*BpfConn) Read ¶

func (bc *BpfConn) Read(b []byte) (n int, err error)

func (*BpfConn) RemoteAddr ¶

func (bc *BpfConn) RemoteAddr() net.Addr

func (*BpfConn) SetDeadline ¶

func (bc *BpfConn) SetDeadline(t time.Time) error

func (*BpfConn) SetReadDeadline ¶

func (bc *BpfConn) SetReadDeadline(t time.Time) error

func (*BpfConn) SetWriteDeadline ¶

func (bc *BpfConn) SetWriteDeadline(t time.Time) error

func (*BpfConn) Write ¶

func (bc *BpfConn) Write(b []byte) (n int, err error)

func (*BpfCount) Attach ¶

func (bc *BpfCount) Attach(conn net.Conn, id uuid.UUID, addr string) (net.Conn, error)

func (*BpfCount) AttachmentState ¶ added in v0.3.0

func (bc *BpfCount) AttachmentState() string

AttachmentState reads the two owned TCX links, without sending packets or changing their configuration. Link.Info can succeed after detach: the kernel then reports ifindex zero. Presence requires the expected hook and program on the configured interface. It proves attachment at observation time, not that traffic is policed. Sessions finish before the node closes these handles.

func (*BpfCount) Close ¶

func (bc *BpfCount) Close() error

func (*BpfCount) DeleteExecLimit ¶

func (bc *BpfCount) DeleteExecLimit(id uuid.UUID) error

func (*BpfCount) DeleteLimit ¶

func (bc *BpfCount) DeleteLimit(addr netutil.IPv6, id uuid.UUID) error

func (*BpfCount) Detach ¶

func (bc *BpfCount) Detach(addr string, id uuid.UUID, ipv6 netutil.IPv6) error

func (*BpfCount) DropTotals ¶ added in v0.3.0

func (bc *BpfCount) DropTotals() (verdicts, skbBytes [2]uint64, err error)

DropTotals reads cumulative observations for this counter instance, ordered ingress then egress. Verdicts count TCX_DROP returns; bytes sum skb->len at those returns. Neither is an on-wire packet/byte count or proof that every traffic path passed through this program. A new counter starts at zero. Sessions finish before the node closes the counter and its owned maps.

func (*BpfCount) SetExecLimit ¶

func (bc *BpfCount) SetExecLimit(id uuid.UUID, limit bitrate.Bitrate) error

func (*BpfCount) SetLimit ¶

func (bc *BpfCount) SetLimit(addr string, id uuid.UUID, limit bitrate.Bitrate) error

func (*BpfCount) Type ¶

func (f *BpfCount) Type() string