introduce std.crypto.CertificateBundle

for reading root certificate authority bundles from standard
installation locations on the file system. So far only Linux logic is
added.

Author: andrewrk

lib/std/crypto.zig

@@ -177,6 +177,8 @@ const std = @import("std.zig");
 pub const errors = @import("crypto/errors.zig");
 
 pub const tls = @import("crypto/tls.zig");
+pub const Der = @import("crypto/Der.zig");
+pub const CertificateBundle = @import("crypto/CertificateBundle.zig");
 
 test {
     _ = aead.aegis.Aegis128L;
@@ -266,6 +268,9 @@ test {
     _ = utils;
     _ = random;
     _ = errors;
+    _ = tls;
+    _ = Der;
+    _ = CertificateBundle;
 }
 
 test "CSPRNG" {

lib/std/crypto/CertificateBundle.zig

@@ -0,0 +1,173 @@
+//! A set of certificates. Typically pre-installed on every operating system,
+//! these are "Certificate Authorities" used to validate SSL certificates.
+//! This data structure stores certificates in DER-encoded form, all of them
+//! concatenated together in the `bytes` array. The `map` field contains an
+//! index from the DER-encoded subject name to the index within `bytes`.
+
+map: std.HashMapUnmanaged(Key, u32, MapContext, std.hash_map.default_max_load_percentage) = .{},
+bytes: std.ArrayListUnmanaged(u8) = .{},
+
+pub const Key = struct {
+    subject_start: u32,
+    subject_end: u32,
+};
+
+/// The returned bytes become invalid after calling any of the rescan functions
+/// or add functions.
+pub fn find(cb: CertificateBundle, subject_name: []const u8) ?[]const u8 {
+    const Adapter = struct {
+        cb: CertificateBundle,
+
+        pub fn hash(ctx: @This(), k: []const u8) u64 {
+            _ = ctx;
+            return std.hash_map.hashString(k);
+        }
+
+        pub fn eql(ctx: @This(), a: []const u8, b_key: Key) bool {
+            const b = ctx.cb.bytes.items[b_key.subject_start..b_key.subject_end];
+            return mem.eql(u8, a, b);
+        }
+    };
+    const index = cb.map.getAdapted(subject_name, Adapter{ .cb = cb }) orelse return null;
+    return cb.bytes.items[index..];
+}
+
+pub fn deinit(cb: *CertificateBundle, gpa: Allocator) void {
+    cb.map.deinit(gpa);
+    cb.bytes.deinit(gpa);
+    cb.* = undefined;
+}
+
+/// Empties the set of certificates and then scans the host operating system
+/// file system standard locations for certificates.
+pub fn rescan(cb: *CertificateBundle, gpa: Allocator) !void {
+    switch (builtin.os.tag) {
+        .linux => return rescanLinux(cb, gpa),
+        else => @compileError("it is unknown where the root CA certificates live on this OS"),
+    }
+}
+
+pub fn rescanLinux(cb: *CertificateBundle, gpa: Allocator) !void {
+    var dir = fs.openIterableDirAbsolute("/etc/ssl/certs", .{}) catch |err| switch (err) {
+        error.FileNotFound => return,
+        else => |e| return e,
+    };
+    defer dir.close();
+
+    cb.bytes.clearRetainingCapacity();
+    cb.map.clearRetainingCapacity();
+
+    var it = dir.iterate();
+    while (try it.next()) |entry| {
+        switch (entry.kind) {
+            .File, .SymLink => {},
+            else => continue,
+        }
+
+        try addCertsFromFile(cb, gpa, dir.dir, entry.name);
+    }
+
+    cb.bytes.shrinkAndFree(gpa, cb.bytes.items.len);
+}
+
+pub fn addCertsFromFile(
+    cb: *CertificateBundle,
+    gpa: Allocator,
+    dir: fs.Dir,
+    sub_file_path: []const u8,
+) !void {
+    var file = try dir.openFile(sub_file_path, .{});
+    defer file.close();
+
+    const size = try file.getEndPos();
+
+    // We borrow `bytes` as a temporary buffer for the base64-encoded data.
+    // This is possible by computing the decoded length and reserving the space
+    // for the decoded bytes first.
+    const decoded_size_upper_bound = size / 4 * 3;
+    try cb.bytes.ensureUnusedCapacity(gpa, decoded_size_upper_bound + size);
+    const end_reserved = cb.bytes.items.len + decoded_size_upper_bound;
+    const buffer = cb.bytes.allocatedSlice()[end_reserved..];
+    const end_index = try file.readAll(buffer);
+    const encoded_bytes = buffer[0..end_index];
+
+    const begin_marker = "-----BEGIN CERTIFICATE-----";
+    const end_marker = "-----END CERTIFICATE-----";
+
+    var start_index: usize = 0;
+    while (mem.indexOfPos(u8, encoded_bytes, start_index, begin_marker)) |begin_marker_start| {
+        const cert_start = begin_marker_start + begin_marker.len;
+        const cert_end = mem.indexOfPos(u8, encoded_bytes, cert_start, end_marker) orelse
+            return error.MissingEndCertificateMarker;
+        start_index = cert_end + end_marker.len;
+        const encoded_cert = mem.trim(u8, encoded_bytes[cert_start..cert_end], " \t\r\n");
+        const decoded_start = @intCast(u32, cb.bytes.items.len);
+        const dest_buf = cb.bytes.allocatedSlice()[decoded_start..];
+        cb.bytes.items.len += try base64.decode(dest_buf, encoded_cert);
+        const k = try key(cb, decoded_start);
+        try cb.map.putContext(gpa, k, decoded_start, .{ .cb = cb });
+    }
+}
+
+pub fn key(cb: *CertificateBundle, bytes_index: u32) !Key {
+    const bytes = cb.bytes.items;
+    const certificate = try Der.parseElement(bytes, bytes_index);
+    const tbs_certificate = try Der.parseElement(bytes, certificate.start);
+    const version = try Der.parseElement(bytes, tbs_certificate.start);
+    if (@bitCast(u8, version.identifier) != 0xa0 or
+        !mem.eql(u8, bytes[version.start..version.end], "\x02\x01\x02"))
+    {
+        return error.UnsupportedCertificateVersion;
+    }
+
+    const serial_number = try Der.parseElement(bytes, version.end);
+
+    // RFC 5280, section 4.1.2.3:
+    // "This field MUST contain the same algorithm identifier as
+    // the signatureAlgorithm field in the sequence Certificate."
+    const signature = try Der.parseElement(bytes, serial_number.end);
+    const issuer = try Der.parseElement(bytes, signature.end);
+    const validity = try Der.parseElement(bytes, issuer.end);
+    const subject = try Der.parseElement(bytes, validity.end);
+    //const subject_pub_key = try Der.parseElement(bytes, subject.end);
+    //const extensions = try Der.parseElement(bytes, subject_pub_key.end);
+
+    return .{
+        .subject_start = subject.start,
+        .subject_end = subject.end,
+    };
+}
+
+const builtin = @import("builtin");
+const std = @import("../std.zig");
+const fs = std.fs;
+const mem = std.mem;
+const Allocator = std.mem.Allocator;
+const Der = std.crypto.Der;
+const CertificateBundle = @This();
+
+const base64 = std.base64.standard.decoderWithIgnore(" \t\r\n");
+
+const MapContext = struct {
+    cb: *const CertificateBundle,
+
+    pub fn hash(ctx: MapContext, k: Key) u64 {
+        return std.hash_map.hashString(ctx.cb.bytes.items[k.subject_start..k.subject_end]);
+    }
+
+    pub fn eql(ctx: MapContext, a: Key, b: Key) bool {
+        const bytes = ctx.cb.bytes.items;
+        return mem.eql(
+            u8,
+            bytes[a.subject_start..a.subject_end],
+            bytes[b.subject_start..b.subject_end],
+        );
+    }
+};
+
+test {
+    var bundle: CertificateBundle = .{};
+    defer bundle.deinit(std.testing.allocator);
+
+    try bundle.rescan(std.testing.allocator);
+}

lib/std/crypto/Der.zig

@@ -0,0 +1,153 @@
+pub const Class = enum(u2) {
+    universal,
+    application,
+    context_specific,
+    private,
+};
+
+pub const PC = enum(u1) {
+    primitive,
+    constructed,
+};
+
+pub const Identifier = packed struct(u8) {
+    tag: Tag,
+    pc: PC,
+    class: Class,
+};
+
+pub const Tag = enum(u5) {
+    boolean = 1,
+    integer = 2,
+    bitstring = 3,
+    null = 5,
+    object_identifier = 6,
+    sequence = 16,
+    sequence_of = 17,
+    _,
+};
+
+pub const Oid = enum {
+    rsadsi,
+    pkcs,
+    rsaEncryption,
+    md2WithRSAEncryption,
+    md5WithRSAEncryption,
+    sha1WithRSAEncryption,
+    sha256WithRSAEncryption,
+    sha384WithRSAEncryption,
+    sha512WithRSAEncryption,
+    sha224WithRSAEncryption,
+    pbeWithMD2AndDES_CBC,
+    pbeWithMD5AndDES_CBC,
+    pkcs9_emailAddress,
+    md2,
+    md5,
+    rc4,
+    ecdsa_with_Recommended,
+    ecdsa_with_Specified,
+    ecdsa_with_SHA224,
+    ecdsa_with_SHA256,
+    ecdsa_with_SHA384,
+    ecdsa_with_SHA512,
+    X500,
+    X509,
+    commonName,
+    serialNumber,
+    countryName,
+    localityName,
+    stateOrProvinceName,
+    organizationName,
+    organizationalUnitName,
+    organizationIdentifier,
+
+    pub const map = std.ComptimeStringMap(Oid, .{
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0x86, 0xF7, 0x0D }, .rsadsi },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0x86, 0xF7, 0x0D, 0x01 }, .pkcs },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0x86, 0xF7, 0x0D, 0x01, 0x01, 0x01 }, .rsaEncryption },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0x86, 0xF7, 0x0D, 0x01, 0x01, 0x02 }, .md2WithRSAEncryption },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0x86, 0xF7, 0x0D, 0x01, 0x01, 0x04 }, .md5WithRSAEncryption },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0x86, 0xF7, 0x0D, 0x01, 0x01, 0x05 }, .sha1WithRSAEncryption },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0x86, 0xF7, 0x0D, 0x01, 0x01, 0x0B }, .sha256WithRSAEncryption },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0x86, 0xF7, 0x0D, 0x01, 0x01, 0x0C }, .sha384WithRSAEncryption },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0x86, 0xF7, 0x0D, 0x01, 0x01, 0x0D }, .sha512WithRSAEncryption },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0x86, 0xF7, 0x0D, 0x01, 0x01, 0x0E }, .sha224WithRSAEncryption },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0x86, 0xF7, 0x0D, 0x01, 0x05, 0x01 }, .pbeWithMD2AndDES_CBC },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0x86, 0xF7, 0x0D, 0x01, 0x05, 0x03 }, .pbeWithMD5AndDES_CBC },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0x86, 0xF7, 0x0D, 0x01, 0x09, 0x01 }, .pkcs9_emailAddress },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0x86, 0xF7, 0x0D, 0x02, 0x02 }, .md2 },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0x86, 0xF7, 0x0D, 0x02, 0x05 }, .md5 },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0x86, 0xF7, 0x0D, 0x03, 0x04 }, .rc4 },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0xCE, 0x3D, 0x04, 0x02 }, .ecdsa_with_Recommended },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0xCE, 0x3D, 0x04, 0x03 }, .ecdsa_with_Specified },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0xCE, 0x3D, 0x04, 0x03, 0x01 }, .ecdsa_with_SHA224 },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0xCE, 0x3D, 0x04, 0x03, 0x02 }, .ecdsa_with_SHA256 },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0xCE, 0x3D, 0x04, 0x03, 0x03 }, .ecdsa_with_SHA384 },
+        .{ &[_]u8{ 0x2A, 0x86, 0x48, 0xCE, 0x3D, 0x04, 0x03, 0x04 }, .ecdsa_with_SHA512 },
+        .{ &[_]u8{0x55}, .X500 },
+        .{ &[_]u8{ 0x55, 0x04 }, .X509 },
+        .{ &[_]u8{ 0x55, 0x04, 0x03 }, .commonName },
+        .{ &[_]u8{ 0x55, 0x04, 0x05 }, .serialNumber },
+        .{ &[_]u8{ 0x55, 0x04, 0x06 }, .countryName },
+        .{ &[_]u8{ 0x55, 0x04, 0x07 }, .localityName },
+        .{ &[_]u8{ 0x55, 0x04, 0x08 }, .stateOrProvinceName },
+        .{ &[_]u8{ 0x55, 0x04, 0x0A }, .organizationName },
+        .{ &[_]u8{ 0x55, 0x04, 0x0B }, .organizationalUnitName },
+        .{ &[_]u8{ 0x55, 0x04, 0x61 }, .organizationIdentifier },
+    });
+};
+
+pub const Element = struct {
+    identifier: Identifier,
+    start: u32,
+    end: u32,
+};
+
+pub const ParseElementError = error{CertificateHasFieldWithInvalidLength};
+
+pub fn parseElement(bytes: []const u8, index: u32) ParseElementError!Element {
+    var i = index;
+    const identifier = @bitCast(Identifier, bytes[i]);
+    i += 1;
+    const size_byte = bytes[i];
+    i += 1;
+    if ((size_byte >> 7) == 0) {
+        return .{
+            .identifier = identifier,
+            .start = i,
+            .end = i + size_byte,
+        };
+    }
+
+    const len_size = @truncate(u7, size_byte);
+    if (len_size > @sizeOf(u32)) {
+        return error.CertificateHasFieldWithInvalidLength;
+    }
+
+    const end_i = i + len_size;
+    var long_form_size: u32 = 0;
+    while (i < end_i) : (i += 1) {
+        long_form_size = (long_form_size << 8) | bytes[i];
+    }
+
+    return .{
+        .identifier = identifier,
+        .start = i,
+        .end = i + long_form_size,
+    };
+}
+
+pub const ParseObjectIdError = error{
+    CertificateHasUnrecognizedObjectId,
+    CertificateFieldHasWrongDataType,
+} || ParseElementError;
+
+pub fn parseObjectId(bytes: []const u8, element: Element) ParseObjectIdError!Oid {
+    if (element.identifier.tag != .object_identifier)
+        return error.CertificateFieldHasWrongDataType;
+    return Oid.map.get(bytes[element.start..element.end]) orelse
+        return error.CertificateHasUnrecognizedObjectId;
+}
+
+const std = @import("../std.zig");
+const Der = @This();