master
  1//! Hexadecimal and Base64 codecs designed for cryptographic use.
  2//! This file provides (best-effort) constant-time encoding and decoding functions for hexadecimal and Base64 formats.
  3//! This is designed to be used in cryptographic applications where timing attacks are a concern.
  4const std = @import("std");
  5const testing = std.testing;
  6const StaticBitSet = std.StaticBitSet;
  7
  8pub const Error = error{
  9    /// An invalid character was found in the input.
 10    InvalidCharacter,
 11    /// The input is not properly padded.
 12    InvalidPadding,
 13    /// The input buffer is too small to hold the output.
 14    NoSpaceLeft,
 15    /// The input and output buffers are not the same size.
 16    SizeMismatch,
 17};
 18
 19/// (best-effort) constant time hexadecimal encoding and decoding.
 20pub const hex = struct {
 21    /// Encodes a binary buffer into a hexadecimal string.
 22    /// The output buffer must be twice the size of the input buffer.
 23    pub fn encode(encoded: []u8, bin: []const u8, comptime case: std.fmt.Case) error{SizeMismatch}!void {
 24        if (encoded.len / 2 != bin.len) {
 25            return error.SizeMismatch;
 26        }
 27        for (bin, 0..) |v, i| {
 28            const b: u16 = v >> 4;
 29            const c: u16 = v & 0xf;
 30            const off = if (case == .upper) 32 else 0;
 31            const x =
 32                ((87 - off + c + (((c -% 10) >> 8) & ~@as(u16, 38 - off))) & 0xff) << 8 |
 33                ((87 - off + b + (((b -% 10) >> 8) & ~@as(u16, 38 - off))) & 0xff);
 34            encoded[i * 2] = @truncate(x);
 35            encoded[i * 2 + 1] = @truncate(x >> 8);
 36        }
 37    }
 38
 39    /// Decodes a hexadecimal string into a binary buffer.
 40    /// The output buffer must be half the size of the input buffer.
 41    pub fn decode(bin: []u8, encoded: []const u8) error{ SizeMismatch, InvalidCharacter, InvalidPadding }!void {
 42        if (encoded.len % 2 != 0) {
 43            return error.InvalidPadding;
 44        }
 45        if (bin.len < encoded.len / 2) {
 46            return error.SizeMismatch;
 47        }
 48        _ = decodeAny(bin, encoded, null) catch |err| {
 49            switch (err) {
 50                error.InvalidCharacter => return error.InvalidCharacter,
 51                error.InvalidPadding => return error.InvalidPadding,
 52                else => unreachable,
 53            }
 54        };
 55    }
 56
 57    /// A decoder that ignores certain characters.
 58    /// The decoder will skip any characters that are in the ignore list.
 59    pub const DecoderWithIgnore = struct {
 60        /// The characters to ignore.
 61        ignored_chars: StaticBitSet(256) = undefined,
 62
 63        /// Decodes a hexadecimal string into a binary buffer.
 64        /// The output buffer must be half the size of the input buffer.
 65        pub fn decode(
 66            self: DecoderWithIgnore,
 67            bin: []u8,
 68            encoded: []const u8,
 69        ) error{ NoSpaceLeft, InvalidCharacter, InvalidPadding }![]const u8 {
 70            return decodeAny(bin, encoded, self.ignored_chars);
 71        }
 72
 73        /// Returns the decoded length of a hexadecimal string, ignoring any characters in the ignore list.
 74        /// This operation does not run in constant time, but it aims to avoid leaking information about the underlying hexadecimal string.
 75        pub fn decodedLenForSlice(decoder: DecoderWithIgnore, encoded: []const u8) !usize {
 76            var hex_len = encoded.len;
 77            for (encoded) |c| {
 78                if (decoder.ignored_chars.isSet(c)) hex_len -= 1;
 79            }
 80            if (hex_len % 2 != 0) {
 81                return error.InvalidPadding;
 82            }
 83            return hex_len / 2;
 84        }
 85
 86        /// Returns the maximum possible decoded size for a given input length after skipping ignored characters.
 87        pub fn decodedLenUpperBound(hex_len: usize) usize {
 88            return hex_len / 2;
 89        }
 90    };
 91
 92    /// Creates a new decoder that ignores certain characters.
 93    /// The decoder will skip any characters that are in the ignore list.
 94    /// The ignore list must not contain any valid hexadecimal characters.
 95    pub fn decoderWithIgnore(ignore_chars: []const u8) error{InvalidCharacter}!DecoderWithIgnore {
 96        var ignored_chars = StaticBitSet(256).initEmpty();
 97        for (ignore_chars) |c| {
 98            switch (c) {
 99                '0'...'9', 'a'...'f', 'A'...'F' => return error.InvalidCharacter,
100                else => if (ignored_chars.isSet(c)) return error.InvalidCharacter,
101            }
102            ignored_chars.set(c);
103        }
104        return DecoderWithIgnore{ .ignored_chars = ignored_chars };
105    }
106
107    fn decodeAny(
108        bin: []u8,
109        encoded: []const u8,
110        ignored_chars: ?StaticBitSet(256),
111    ) error{ NoSpaceLeft, InvalidCharacter, InvalidPadding }![]const u8 {
112        var bin_pos: usize = 0;
113        var state: bool = false;
114        var c_acc: u8 = 0;
115        for (encoded) |c| {
116            const c_num = c ^ 48;
117            const c_num0: u8 = @truncate((@as(u16, c_num) -% 10) >> 8);
118            const c_alpha: u8 = (c & ~@as(u8, 32)) -% 55;
119            const c_alpha0: u8 = @truncate(((@as(u16, c_alpha) -% 10) ^ (@as(u16, c_alpha) -% 16)) >> 8);
120            if ((c_num0 | c_alpha0) == 0) {
121                if (ignored_chars) |set| {
122                    if (set.isSet(c)) {
123                        continue;
124                    }
125                }
126                return error.InvalidCharacter;
127            }
128            const c_val = (c_num0 & c_num) | (c_alpha0 & c_alpha);
129            if (bin_pos >= bin.len) {
130                return error.NoSpaceLeft;
131            }
132            if (!state) {
133                c_acc = c_val << 4;
134            } else {
135                bin[bin_pos] = c_acc | c_val;
136                bin_pos += 1;
137            }
138            state = !state;
139        }
140        if (state) {
141            return error.InvalidPadding;
142        }
143        return bin[0..bin_pos];
144    }
145};
146
147/// (best-effort) constant time base64 encoding and decoding.
148pub const base64 = struct {
149    /// The base64 variant to use.
150    pub const Variant = packed struct {
151        /// Use the URL-safe alphabet instead of the standard alphabet.
152        urlsafe_alphabet: bool = false,
153        /// Enable padding with '=' characters.
154        padding: bool = true,
155
156        /// The standard base64 variant.
157        pub const standard: Variant = .{ .urlsafe_alphabet = false, .padding = true };
158        /// The URL-safe base64 variant.
159        pub const urlsafe: Variant = .{ .urlsafe_alphabet = true, .padding = true };
160        /// The standard base64 variant without padding.
161        pub const standard_nopad: Variant = .{ .urlsafe_alphabet = false, .padding = false };
162        /// The URL-safe base64 variant without padding.
163        pub const urlsafe_nopad: Variant = .{ .urlsafe_alphabet = true, .padding = false };
164    };
165
166    /// Returns the length of the encoded base64 string for a given length.
167    pub fn encodedLen(bin_len: usize, variant: Variant) usize {
168        if (variant.padding) {
169            return (bin_len + 2) / 3 * 4;
170        } else {
171            const leftover = bin_len % 3;
172            return bin_len / 3 * 4 + (leftover * 4 + 2) / 3;
173        }
174    }
175
176    /// Returns the maximum possible decoded size for a given input length - The actual length may be less if the input includes padding.
177    /// `InvalidPadding` is returned if the input length is not valid.
178    pub fn decodedLen(b64_len: usize, variant: Variant) !usize {
179        var result = b64_len / 4 * 3;
180        const leftover = b64_len % 4;
181        if (variant.padding) {
182            if (leftover % 4 != 0) return error.InvalidPadding;
183        } else {
184            if (leftover % 4 == 1) return error.InvalidPadding;
185            result += leftover * 3 / 4;
186        }
187        return result;
188    }
189
190    /// Encodes a binary buffer into a base64 string.
191    /// The output buffer must be at least `encodedLen(bin.len)` bytes long.
192    pub fn encode(encoded: []u8, bin: []const u8, comptime variant: Variant) error{NoSpaceLeft}![]const u8 {
193        var acc_len: u4 = 0;
194        var b64_pos: usize = 0;
195        var acc: u16 = 0;
196        const nibbles = bin.len / 3;
197        const remainder = bin.len - 3 * nibbles;
198        var b64_len = nibbles * 4;
199        if (remainder != 0) {
200            b64_len += if (variant.padding) 4 else 2 + (remainder >> 1);
201        }
202        if (encoded.len < b64_len) {
203            return error.NoSpaceLeft;
204        }
205        const urlsafe = variant.urlsafe_alphabet;
206        for (bin) |v| {
207            acc = (acc << 8) + v;
208            acc_len += 8;
209            while (acc_len >= 6) {
210                acc_len -= 6;
211                encoded[b64_pos] = charFromByte(@as(u6, @truncate(acc >> acc_len)), urlsafe);
212                b64_pos += 1;
213            }
214        }
215        if (acc_len > 0) {
216            encoded[b64_pos] = charFromByte(@as(u6, @truncate(acc << (6 - acc_len))), urlsafe);
217            b64_pos += 1;
218        }
219        while (b64_pos < b64_len) {
220            encoded[b64_pos] = '=';
221            b64_pos += 1;
222        }
223        return encoded[0..b64_pos];
224    }
225
226    /// Decodes a base64 string into a binary buffer.
227    /// The output buffer must be at least `decodedLenUpperBound(encoded.len)` bytes long.
228    pub fn decode(bin: []u8, encoded: []const u8, comptime variant: Variant) error{ InvalidCharacter, InvalidPadding }![]const u8 {
229        return decodeAny(bin, encoded, variant, null) catch |err| {
230            switch (err) {
231                error.InvalidCharacter => return error.InvalidCharacter,
232                error.InvalidPadding => return error.InvalidPadding,
233                else => unreachable,
234            }
235        };
236    }
237
238    //// A decoder that ignores certain characters.
239    pub const DecoderWithIgnore = struct {
240        /// The characters to ignore.
241        ignored_chars: StaticBitSet(256) = undefined,
242
243        /// Decodes a base64 string into a binary buffer.
244        /// The output buffer must be at least `decodedLenUpperBound(encoded.len)` bytes long.
245        pub fn decode(
246            self: DecoderWithIgnore,
247            bin: []u8,
248            encoded: []const u8,
249            comptime variant: Variant,
250        ) error{ NoSpaceLeft, InvalidCharacter, InvalidPadding }![]const u8 {
251            return decodeAny(bin, encoded, variant, self.ignored_chars);
252        }
253
254        /// Returns the decoded length of a base64 string, ignoring any characters in the ignore list.
255        /// This operation does not run in constant time, but it aims to avoid leaking information about the underlying base64 string.
256        pub fn decodedLenForSlice(decoder: DecoderWithIgnore, encoded: []const u8, variant: Variant) !usize {
257            var b64_len = encoded.len;
258            for (encoded) |c| {
259                if (decoder.ignored_chars.isSet(c)) b64_len -= 1;
260            }
261            return base64.decodedLen(b64_len, variant);
262        }
263
264        /// Returns the maximum possible decoded size for a given input length after skipping ignored characters.
265        pub fn decodedLenUpperBound(b64_len: usize) usize {
266            return b64_len / 3 * 4;
267        }
268    };
269
270    /// Creates a new decoder that ignores certain characters.
271    pub fn decoderWithIgnore(ignore_chars: []const u8) error{InvalidCharacter}!DecoderWithIgnore {
272        var ignored_chars = StaticBitSet(256).initEmpty();
273        for (ignore_chars) |c| {
274            switch (c) {
275                'A'...'Z', 'a'...'z', '0'...'9' => return error.InvalidCharacter,
276                else => if (ignored_chars.isSet(c)) return error.InvalidCharacter,
277            }
278            ignored_chars.set(c);
279        }
280        return DecoderWithIgnore{ .ignored_chars = ignored_chars };
281    }
282
283    fn eq(x: u8, y: u8) u8 {
284        return ~@as(u8, @truncate((0 -% (@as(u16, x) ^ @as(u16, y))) >> 8));
285    }
286
287    fn gt(x: u8, y: u8) u8 {
288        return @truncate((@as(u16, y) -% @as(u16, x)) >> 8);
289    }
290
291    fn ge(x: u8, y: u8) u8 {
292        return ~gt(y, x);
293    }
294
295    fn lt(x: u8, y: u8) u8 {
296        return gt(y, x);
297    }
298
299    fn le(x: u8, y: u8) u8 {
300        return ge(y, x);
301    }
302
303    fn charFromByte(x: u8, comptime urlsafe: bool) u8 {
304        return (lt(x, 26) & (x +% 'A')) |
305            (ge(x, 26) & lt(x, 52) & (x +% 'a' -% 26)) |
306            (ge(x, 52) & lt(x, 62) & (x +% '0' -% 52)) |
307            (eq(x, 62) & '+') | (eq(x, 63) & if (urlsafe) '_' else '/');
308    }
309
310    fn byteFromChar(c: u8, comptime urlsafe: bool) u8 {
311        const x =
312            (ge(c, 'A') & le(c, 'Z') & (c -% 'A')) |
313            (ge(c, 'a') & le(c, 'z') & (c -% 'a' +% 26)) |
314            (ge(c, '0') & le(c, '9') & (c -% '0' +% 52)) |
315            (eq(c, '+') & 62) | (eq(c, if (urlsafe) '_' else '/') & 63);
316        return x | (eq(x, 0) & ~eq(c, 'A'));
317    }
318
319    fn skipPadding(
320        encoded: []const u8,
321        padding_len: usize,
322        ignored_chars: ?StaticBitSet(256),
323    ) error{InvalidPadding}![]const u8 {
324        var b64_pos: usize = 0;
325        var i = padding_len;
326        while (i > 0) {
327            if (b64_pos >= encoded.len) {
328                return error.InvalidPadding;
329            }
330            const c = encoded[b64_pos];
331            if (c == '=') {
332                i -= 1;
333            } else if (ignored_chars) |set| {
334                if (!set.isSet(c)) {
335                    return error.InvalidPadding;
336                }
337            }
338            b64_pos += 1;
339        }
340        return encoded[b64_pos..];
341    }
342
343    fn decodeAny(
344        bin: []u8,
345        encoded: []const u8,
346        comptime variant: Variant,
347        ignored_chars: ?StaticBitSet(256),
348    ) error{ NoSpaceLeft, InvalidCharacter, InvalidPadding }![]const u8 {
349        var acc: u16 = 0;
350        var acc_len: u4 = 0;
351        var bin_pos: usize = 0;
352        var premature_end: ?usize = null;
353        const urlsafe = variant.urlsafe_alphabet;
354        for (encoded, 0..) |c, b64_pos| {
355            const d = byteFromChar(c, urlsafe);
356            if (d == 0xff) {
357                if (ignored_chars) |set| {
358                    if (set.isSet(c)) continue;
359                }
360                premature_end = b64_pos;
361                break;
362            }
363            acc = (acc << 6) + d;
364            acc_len += 6;
365            if (acc_len >= 8) {
366                acc_len -= 8;
367                if (bin_pos >= bin.len) {
368                    return error.NoSpaceLeft;
369                }
370                bin[bin_pos] = @truncate(acc >> acc_len);
371                bin_pos += 1;
372            }
373        }
374        if (acc_len > 4 or (acc & ((@as(u16, 1) << acc_len) -% 1)) != 0) {
375            return error.InvalidCharacter;
376        }
377        const padding_len = acc_len / 2;
378        if (premature_end) |pos| {
379            const remaining =
380                if (variant.padding)
381                    try skipPadding(encoded[pos..], padding_len, ignored_chars)
382                else
383                    encoded[pos..];
384            if (ignored_chars) |set| {
385                for (remaining) |c| {
386                    if (!set.isSet(c)) {
387                        return error.InvalidCharacter;
388                    }
389                }
390            } else if (remaining.len != 0) {
391                return error.InvalidCharacter;
392            }
393        } else if (variant.padding and padding_len != 0) {
394            return error.InvalidPadding;
395        }
396        return bin[0..bin_pos];
397    }
398};
399
400test "hex" {
401    var default_rng = std.Random.DefaultPrng.init(testing.random_seed);
402    var rng = default_rng.random();
403    var bin_buf: [1000]u8 = undefined;
404    rng.bytes(&bin_buf);
405    var bin2_buf: [bin_buf.len]u8 = undefined;
406    var hex_buf: [bin_buf.len * 2]u8 = undefined;
407    for (0..1000) |_| {
408        const bin_len = rng.intRangeAtMost(usize, 0, bin_buf.len);
409        const bin = bin_buf[0..bin_len];
410        const bin2 = bin2_buf[0..bin_len];
411        inline for (.{ .lower, .upper }) |case| {
412            const hex_len = bin_len * 2;
413            const encoded = hex_buf[0..hex_len];
414            try hex.encode(encoded, bin, case);
415            try hex.decode(bin2, encoded);
416            try testing.expectEqualSlices(u8, bin, bin2);
417        }
418    }
419}
420
421test "base64" {
422    var default_rng = std.Random.DefaultPrng.init(testing.random_seed);
423    var rng = default_rng.random();
424    var bin_buf: [1000]u8 = undefined;
425    rng.bytes(&bin_buf);
426    var bin2_buf: [bin_buf.len]u8 = undefined;
427    var b64_buf: [(bin_buf.len + 3) / 3 * 4]u8 = undefined;
428    for (0..1000) |_| {
429        const bin_len = rng.intRangeAtMost(usize, 0, bin_buf.len);
430        const bin = bin_buf[0..bin_len];
431        const bin2 = bin2_buf[0..bin_len];
432        inline for ([_]base64.Variant{
433            .standard,
434            .standard_nopad,
435            .urlsafe,
436            .urlsafe_nopad,
437        }) |variant| {
438            const b64_len = base64.encodedLen(bin_len, variant);
439            const encoded_buf = b64_buf[0..b64_len];
440            const encoded = try base64.encode(encoded_buf, bin, variant);
441            const decoded = try base64.decode(bin2, encoded, variant);
442            try testing.expectEqualSlices(u8, bin, decoded);
443        }
444    }
445}
446
447test "hex with ignored chars" {
448    const encoded = "01020304050607\n08090A0B0C0D0E0F\n";
449    const expected = [_]u8{ 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x09, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, 0x0F };
450    var bin_buf: [encoded.len / 2]u8 = undefined;
451    try testing.expectError(error.InvalidCharacter, hex.decode(&bin_buf, encoded));
452    const bin = try (try hex.decoderWithIgnore("\r\n")).decode(&bin_buf, encoded);
453    try testing.expectEqualSlices(u8, &expected, bin);
454}
455
456test "base64 with ignored chars" {
457    const encoded = "dGVzdCBi\r\nYXNlNjQ=\n";
458    const expected = "test base64";
459    var bin_buf: [base64.DecoderWithIgnore.decodedLenUpperBound(encoded.len)]u8 = undefined;
460    try testing.expectError(error.InvalidCharacter, base64.decode(&bin_buf, encoded, .standard));
461    const bin = try (try base64.decoderWithIgnore("\r\n")).decode(&bin_buf, encoded, .standard);
462    try testing.expectEqualSlices(u8, expected, bin);
463}