mirror of
https://github.com/ceph/ceph-csi.git
synced 2024-12-21 04:20:23 +00:00
91774fc936
Uses github.com/libopenstorage/secrets to communicate with Vault. This removes the need for maintaining our own limited Vault APIs. By adding the new dependency, several other packages got updated in the process. Unused indirect dependencies have been removed from go.mod. Signed-off-by: Niels de Vos <ndevos@redhat.com>
539 lines
13 KiB
Go
539 lines
13 KiB
Go
// Copyright 2018 The Go Authors. All rights reserved.
|
|
// Use of this source code is governed by a BSD-style
|
|
// license that can be found in the LICENSE file.
|
|
|
|
// Package protowire parses and formats the raw wire encoding.
|
|
// See https://developers.google.com/protocol-buffers/docs/encoding.
|
|
//
|
|
// For marshaling and unmarshaling entire protobuf messages,
|
|
// use the "google.golang.org/protobuf/proto" package instead.
|
|
package protowire
|
|
|
|
import (
|
|
"io"
|
|
"math"
|
|
"math/bits"
|
|
|
|
"google.golang.org/protobuf/internal/errors"
|
|
)
|
|
|
|
// Number represents the field number.
|
|
type Number int32
|
|
|
|
const (
|
|
MinValidNumber Number = 1
|
|
FirstReservedNumber Number = 19000
|
|
LastReservedNumber Number = 19999
|
|
MaxValidNumber Number = 1<<29 - 1
|
|
)
|
|
|
|
// IsValid reports whether the field number is semantically valid.
|
|
//
|
|
// Note that while numbers within the reserved range are semantically invalid,
|
|
// they are syntactically valid in the wire format.
|
|
// Implementations may treat records with reserved field numbers as unknown.
|
|
func (n Number) IsValid() bool {
|
|
return MinValidNumber <= n && n < FirstReservedNumber || LastReservedNumber < n && n <= MaxValidNumber
|
|
}
|
|
|
|
// Type represents the wire type.
|
|
type Type int8
|
|
|
|
const (
|
|
VarintType Type = 0
|
|
Fixed32Type Type = 5
|
|
Fixed64Type Type = 1
|
|
BytesType Type = 2
|
|
StartGroupType Type = 3
|
|
EndGroupType Type = 4
|
|
)
|
|
|
|
const (
|
|
_ = -iota
|
|
errCodeTruncated
|
|
errCodeFieldNumber
|
|
errCodeOverflow
|
|
errCodeReserved
|
|
errCodeEndGroup
|
|
)
|
|
|
|
var (
|
|
errFieldNumber = errors.New("invalid field number")
|
|
errOverflow = errors.New("variable length integer overflow")
|
|
errReserved = errors.New("cannot parse reserved wire type")
|
|
errEndGroup = errors.New("mismatching end group marker")
|
|
errParse = errors.New("parse error")
|
|
)
|
|
|
|
// ParseError converts an error code into an error value.
|
|
// This returns nil if n is a non-negative number.
|
|
func ParseError(n int) error {
|
|
if n >= 0 {
|
|
return nil
|
|
}
|
|
switch n {
|
|
case errCodeTruncated:
|
|
return io.ErrUnexpectedEOF
|
|
case errCodeFieldNumber:
|
|
return errFieldNumber
|
|
case errCodeOverflow:
|
|
return errOverflow
|
|
case errCodeReserved:
|
|
return errReserved
|
|
case errCodeEndGroup:
|
|
return errEndGroup
|
|
default:
|
|
return errParse
|
|
}
|
|
}
|
|
|
|
// ConsumeField parses an entire field record (both tag and value) and returns
|
|
// the field number, the wire type, and the total length.
|
|
// This returns a negative length upon an error (see ParseError).
|
|
//
|
|
// The total length includes the tag header and the end group marker (if the
|
|
// field is a group).
|
|
func ConsumeField(b []byte) (Number, Type, int) {
|
|
num, typ, n := ConsumeTag(b)
|
|
if n < 0 {
|
|
return 0, 0, n // forward error code
|
|
}
|
|
m := ConsumeFieldValue(num, typ, b[n:])
|
|
if m < 0 {
|
|
return 0, 0, m // forward error code
|
|
}
|
|
return num, typ, n + m
|
|
}
|
|
|
|
// ConsumeFieldValue parses a field value and returns its length.
|
|
// This assumes that the field Number and wire Type have already been parsed.
|
|
// This returns a negative length upon an error (see ParseError).
|
|
//
|
|
// When parsing a group, the length includes the end group marker and
|
|
// the end group is verified to match the starting field number.
|
|
func ConsumeFieldValue(num Number, typ Type, b []byte) (n int) {
|
|
switch typ {
|
|
case VarintType:
|
|
_, n = ConsumeVarint(b)
|
|
return n
|
|
case Fixed32Type:
|
|
_, n = ConsumeFixed32(b)
|
|
return n
|
|
case Fixed64Type:
|
|
_, n = ConsumeFixed64(b)
|
|
return n
|
|
case BytesType:
|
|
_, n = ConsumeBytes(b)
|
|
return n
|
|
case StartGroupType:
|
|
n0 := len(b)
|
|
for {
|
|
num2, typ2, n := ConsumeTag(b)
|
|
if n < 0 {
|
|
return n // forward error code
|
|
}
|
|
b = b[n:]
|
|
if typ2 == EndGroupType {
|
|
if num != num2 {
|
|
return errCodeEndGroup
|
|
}
|
|
return n0 - len(b)
|
|
}
|
|
|
|
n = ConsumeFieldValue(num2, typ2, b)
|
|
if n < 0 {
|
|
return n // forward error code
|
|
}
|
|
b = b[n:]
|
|
}
|
|
case EndGroupType:
|
|
return errCodeEndGroup
|
|
default:
|
|
return errCodeReserved
|
|
}
|
|
}
|
|
|
|
// AppendTag encodes num and typ as a varint-encoded tag and appends it to b.
|
|
func AppendTag(b []byte, num Number, typ Type) []byte {
|
|
return AppendVarint(b, EncodeTag(num, typ))
|
|
}
|
|
|
|
// ConsumeTag parses b as a varint-encoded tag, reporting its length.
|
|
// This returns a negative length upon an error (see ParseError).
|
|
func ConsumeTag(b []byte) (Number, Type, int) {
|
|
v, n := ConsumeVarint(b)
|
|
if n < 0 {
|
|
return 0, 0, n // forward error code
|
|
}
|
|
num, typ := DecodeTag(v)
|
|
if num < MinValidNumber {
|
|
return 0, 0, errCodeFieldNumber
|
|
}
|
|
return num, typ, n
|
|
}
|
|
|
|
func SizeTag(num Number) int {
|
|
return SizeVarint(EncodeTag(num, 0)) // wire type has no effect on size
|
|
}
|
|
|
|
// AppendVarint appends v to b as a varint-encoded uint64.
|
|
func AppendVarint(b []byte, v uint64) []byte {
|
|
switch {
|
|
case v < 1<<7:
|
|
b = append(b, byte(v))
|
|
case v < 1<<14:
|
|
b = append(b,
|
|
byte((v>>0)&0x7f|0x80),
|
|
byte(v>>7))
|
|
case v < 1<<21:
|
|
b = append(b,
|
|
byte((v>>0)&0x7f|0x80),
|
|
byte((v>>7)&0x7f|0x80),
|
|
byte(v>>14))
|
|
case v < 1<<28:
|
|
b = append(b,
|
|
byte((v>>0)&0x7f|0x80),
|
|
byte((v>>7)&0x7f|0x80),
|
|
byte((v>>14)&0x7f|0x80),
|
|
byte(v>>21))
|
|
case v < 1<<35:
|
|
b = append(b,
|
|
byte((v>>0)&0x7f|0x80),
|
|
byte((v>>7)&0x7f|0x80),
|
|
byte((v>>14)&0x7f|0x80),
|
|
byte((v>>21)&0x7f|0x80),
|
|
byte(v>>28))
|
|
case v < 1<<42:
|
|
b = append(b,
|
|
byte((v>>0)&0x7f|0x80),
|
|
byte((v>>7)&0x7f|0x80),
|
|
byte((v>>14)&0x7f|0x80),
|
|
byte((v>>21)&0x7f|0x80),
|
|
byte((v>>28)&0x7f|0x80),
|
|
byte(v>>35))
|
|
case v < 1<<49:
|
|
b = append(b,
|
|
byte((v>>0)&0x7f|0x80),
|
|
byte((v>>7)&0x7f|0x80),
|
|
byte((v>>14)&0x7f|0x80),
|
|
byte((v>>21)&0x7f|0x80),
|
|
byte((v>>28)&0x7f|0x80),
|
|
byte((v>>35)&0x7f|0x80),
|
|
byte(v>>42))
|
|
case v < 1<<56:
|
|
b = append(b,
|
|
byte((v>>0)&0x7f|0x80),
|
|
byte((v>>7)&0x7f|0x80),
|
|
byte((v>>14)&0x7f|0x80),
|
|
byte((v>>21)&0x7f|0x80),
|
|
byte((v>>28)&0x7f|0x80),
|
|
byte((v>>35)&0x7f|0x80),
|
|
byte((v>>42)&0x7f|0x80),
|
|
byte(v>>49))
|
|
case v < 1<<63:
|
|
b = append(b,
|
|
byte((v>>0)&0x7f|0x80),
|
|
byte((v>>7)&0x7f|0x80),
|
|
byte((v>>14)&0x7f|0x80),
|
|
byte((v>>21)&0x7f|0x80),
|
|
byte((v>>28)&0x7f|0x80),
|
|
byte((v>>35)&0x7f|0x80),
|
|
byte((v>>42)&0x7f|0x80),
|
|
byte((v>>49)&0x7f|0x80),
|
|
byte(v>>56))
|
|
default:
|
|
b = append(b,
|
|
byte((v>>0)&0x7f|0x80),
|
|
byte((v>>7)&0x7f|0x80),
|
|
byte((v>>14)&0x7f|0x80),
|
|
byte((v>>21)&0x7f|0x80),
|
|
byte((v>>28)&0x7f|0x80),
|
|
byte((v>>35)&0x7f|0x80),
|
|
byte((v>>42)&0x7f|0x80),
|
|
byte((v>>49)&0x7f|0x80),
|
|
byte((v>>56)&0x7f|0x80),
|
|
1)
|
|
}
|
|
return b
|
|
}
|
|
|
|
// ConsumeVarint parses b as a varint-encoded uint64, reporting its length.
|
|
// This returns a negative length upon an error (see ParseError).
|
|
func ConsumeVarint(b []byte) (v uint64, n int) {
|
|
var y uint64
|
|
if len(b) <= 0 {
|
|
return 0, errCodeTruncated
|
|
}
|
|
v = uint64(b[0])
|
|
if v < 0x80 {
|
|
return v, 1
|
|
}
|
|
v -= 0x80
|
|
|
|
if len(b) <= 1 {
|
|
return 0, errCodeTruncated
|
|
}
|
|
y = uint64(b[1])
|
|
v += y << 7
|
|
if y < 0x80 {
|
|
return v, 2
|
|
}
|
|
v -= 0x80 << 7
|
|
|
|
if len(b) <= 2 {
|
|
return 0, errCodeTruncated
|
|
}
|
|
y = uint64(b[2])
|
|
v += y << 14
|
|
if y < 0x80 {
|
|
return v, 3
|
|
}
|
|
v -= 0x80 << 14
|
|
|
|
if len(b) <= 3 {
|
|
return 0, errCodeTruncated
|
|
}
|
|
y = uint64(b[3])
|
|
v += y << 21
|
|
if y < 0x80 {
|
|
return v, 4
|
|
}
|
|
v -= 0x80 << 21
|
|
|
|
if len(b) <= 4 {
|
|
return 0, errCodeTruncated
|
|
}
|
|
y = uint64(b[4])
|
|
v += y << 28
|
|
if y < 0x80 {
|
|
return v, 5
|
|
}
|
|
v -= 0x80 << 28
|
|
|
|
if len(b) <= 5 {
|
|
return 0, errCodeTruncated
|
|
}
|
|
y = uint64(b[5])
|
|
v += y << 35
|
|
if y < 0x80 {
|
|
return v, 6
|
|
}
|
|
v -= 0x80 << 35
|
|
|
|
if len(b) <= 6 {
|
|
return 0, errCodeTruncated
|
|
}
|
|
y = uint64(b[6])
|
|
v += y << 42
|
|
if y < 0x80 {
|
|
return v, 7
|
|
}
|
|
v -= 0x80 << 42
|
|
|
|
if len(b) <= 7 {
|
|
return 0, errCodeTruncated
|
|
}
|
|
y = uint64(b[7])
|
|
v += y << 49
|
|
if y < 0x80 {
|
|
return v, 8
|
|
}
|
|
v -= 0x80 << 49
|
|
|
|
if len(b) <= 8 {
|
|
return 0, errCodeTruncated
|
|
}
|
|
y = uint64(b[8])
|
|
v += y << 56
|
|
if y < 0x80 {
|
|
return v, 9
|
|
}
|
|
v -= 0x80 << 56
|
|
|
|
if len(b) <= 9 {
|
|
return 0, errCodeTruncated
|
|
}
|
|
y = uint64(b[9])
|
|
v += y << 63
|
|
if y < 2 {
|
|
return v, 10
|
|
}
|
|
return 0, errCodeOverflow
|
|
}
|
|
|
|
// SizeVarint returns the encoded size of a varint.
|
|
// The size is guaranteed to be within 1 and 10, inclusive.
|
|
func SizeVarint(v uint64) int {
|
|
// This computes 1 + (bits.Len64(v)-1)/7.
|
|
// 9/64 is a good enough approximation of 1/7
|
|
return int(9*uint32(bits.Len64(v))+64) / 64
|
|
}
|
|
|
|
// AppendFixed32 appends v to b as a little-endian uint32.
|
|
func AppendFixed32(b []byte, v uint32) []byte {
|
|
return append(b,
|
|
byte(v>>0),
|
|
byte(v>>8),
|
|
byte(v>>16),
|
|
byte(v>>24))
|
|
}
|
|
|
|
// ConsumeFixed32 parses b as a little-endian uint32, reporting its length.
|
|
// This returns a negative length upon an error (see ParseError).
|
|
func ConsumeFixed32(b []byte) (v uint32, n int) {
|
|
if len(b) < 4 {
|
|
return 0, errCodeTruncated
|
|
}
|
|
v = uint32(b[0])<<0 | uint32(b[1])<<8 | uint32(b[2])<<16 | uint32(b[3])<<24
|
|
return v, 4
|
|
}
|
|
|
|
// SizeFixed32 returns the encoded size of a fixed32; which is always 4.
|
|
func SizeFixed32() int {
|
|
return 4
|
|
}
|
|
|
|
// AppendFixed64 appends v to b as a little-endian uint64.
|
|
func AppendFixed64(b []byte, v uint64) []byte {
|
|
return append(b,
|
|
byte(v>>0),
|
|
byte(v>>8),
|
|
byte(v>>16),
|
|
byte(v>>24),
|
|
byte(v>>32),
|
|
byte(v>>40),
|
|
byte(v>>48),
|
|
byte(v>>56))
|
|
}
|
|
|
|
// ConsumeFixed64 parses b as a little-endian uint64, reporting its length.
|
|
// This returns a negative length upon an error (see ParseError).
|
|
func ConsumeFixed64(b []byte) (v uint64, n int) {
|
|
if len(b) < 8 {
|
|
return 0, errCodeTruncated
|
|
}
|
|
v = uint64(b[0])<<0 | uint64(b[1])<<8 | uint64(b[2])<<16 | uint64(b[3])<<24 | uint64(b[4])<<32 | uint64(b[5])<<40 | uint64(b[6])<<48 | uint64(b[7])<<56
|
|
return v, 8
|
|
}
|
|
|
|
// SizeFixed64 returns the encoded size of a fixed64; which is always 8.
|
|
func SizeFixed64() int {
|
|
return 8
|
|
}
|
|
|
|
// AppendBytes appends v to b as a length-prefixed bytes value.
|
|
func AppendBytes(b []byte, v []byte) []byte {
|
|
return append(AppendVarint(b, uint64(len(v))), v...)
|
|
}
|
|
|
|
// ConsumeBytes parses b as a length-prefixed bytes value, reporting its length.
|
|
// This returns a negative length upon an error (see ParseError).
|
|
func ConsumeBytes(b []byte) (v []byte, n int) {
|
|
m, n := ConsumeVarint(b)
|
|
if n < 0 {
|
|
return nil, n // forward error code
|
|
}
|
|
if m > uint64(len(b[n:])) {
|
|
return nil, errCodeTruncated
|
|
}
|
|
return b[n:][:m], n + int(m)
|
|
}
|
|
|
|
// SizeBytes returns the encoded size of a length-prefixed bytes value,
|
|
// given only the length.
|
|
func SizeBytes(n int) int {
|
|
return SizeVarint(uint64(n)) + n
|
|
}
|
|
|
|
// AppendString appends v to b as a length-prefixed bytes value.
|
|
func AppendString(b []byte, v string) []byte {
|
|
return append(AppendVarint(b, uint64(len(v))), v...)
|
|
}
|
|
|
|
// ConsumeString parses b as a length-prefixed bytes value, reporting its length.
|
|
// This returns a negative length upon an error (see ParseError).
|
|
func ConsumeString(b []byte) (v string, n int) {
|
|
bb, n := ConsumeBytes(b)
|
|
return string(bb), n
|
|
}
|
|
|
|
// AppendGroup appends v to b as group value, with a trailing end group marker.
|
|
// The value v must not contain the end marker.
|
|
func AppendGroup(b []byte, num Number, v []byte) []byte {
|
|
return AppendVarint(append(b, v...), EncodeTag(num, EndGroupType))
|
|
}
|
|
|
|
// ConsumeGroup parses b as a group value until the trailing end group marker,
|
|
// and verifies that the end marker matches the provided num. The value v
|
|
// does not contain the end marker, while the length does contain the end marker.
|
|
// This returns a negative length upon an error (see ParseError).
|
|
func ConsumeGroup(num Number, b []byte) (v []byte, n int) {
|
|
n = ConsumeFieldValue(num, StartGroupType, b)
|
|
if n < 0 {
|
|
return nil, n // forward error code
|
|
}
|
|
b = b[:n]
|
|
|
|
// Truncate off end group marker, but need to handle denormalized varints.
|
|
// Assuming end marker is never 0 (which is always the case since
|
|
// EndGroupType is non-zero), we can truncate all trailing bytes where the
|
|
// lower 7 bits are all zero (implying that the varint is denormalized).
|
|
for len(b) > 0 && b[len(b)-1]&0x7f == 0 {
|
|
b = b[:len(b)-1]
|
|
}
|
|
b = b[:len(b)-SizeTag(num)]
|
|
return b, n
|
|
}
|
|
|
|
// SizeGroup returns the encoded size of a group, given only the length.
|
|
func SizeGroup(num Number, n int) int {
|
|
return n + SizeTag(num)
|
|
}
|
|
|
|
// DecodeTag decodes the field Number and wire Type from its unified form.
|
|
// The Number is -1 if the decoded field number overflows int32.
|
|
// Other than overflow, this does not check for field number validity.
|
|
func DecodeTag(x uint64) (Number, Type) {
|
|
// NOTE: MessageSet allows for larger field numbers than normal.
|
|
if x>>3 > uint64(math.MaxInt32) {
|
|
return -1, 0
|
|
}
|
|
return Number(x >> 3), Type(x & 7)
|
|
}
|
|
|
|
// EncodeTag encodes the field Number and wire Type into its unified form.
|
|
func EncodeTag(num Number, typ Type) uint64 {
|
|
return uint64(num)<<3 | uint64(typ&7)
|
|
}
|
|
|
|
// DecodeZigZag decodes a zig-zag-encoded uint64 as an int64.
|
|
// Input: {…, 5, 3, 1, 0, 2, 4, 6, …}
|
|
// Output: {…, -3, -2, -1, 0, +1, +2, +3, …}
|
|
func DecodeZigZag(x uint64) int64 {
|
|
return int64(x>>1) ^ int64(x)<<63>>63
|
|
}
|
|
|
|
// EncodeZigZag encodes an int64 as a zig-zag-encoded uint64.
|
|
// Input: {…, -3, -2, -1, 0, +1, +2, +3, …}
|
|
// Output: {…, 5, 3, 1, 0, 2, 4, 6, …}
|
|
func EncodeZigZag(x int64) uint64 {
|
|
return uint64(x<<1) ^ uint64(x>>63)
|
|
}
|
|
|
|
// DecodeBool decodes a uint64 as a bool.
|
|
// Input: { 0, 1, 2, …}
|
|
// Output: {false, true, true, …}
|
|
func DecodeBool(x uint64) bool {
|
|
return x != 0
|
|
}
|
|
|
|
// EncodeBool encodes a bool as a uint64.
|
|
// Input: {false, true}
|
|
// Output: { 0, 1}
|
|
func EncodeBool(x bool) uint64 {
|
|
if x {
|
|
return 1
|
|
}
|
|
return 0
|
|
}
|