Function utf8_sequence_len

src/os/linux/pagecache.cpp:98–124

Length (1-4) of a well-formed UTF-8 sequence starting at p with `avail` bytes remaining, or 0 if the bytes are not well-formed UTF-8. Follows the Unicode 3.9 well-formed byte-sequence table: rejects overlong encodings, UTF-16 surrogates (U+D800–U+DFFF), and codepoints above U+10FFFF, as well as lone continuation bytes and truncated sequences. This is what tells a real non-ASCII filename apart from

// lone continuation bytes and truncated sequences. This is what tells a real
// non-ASCII filename apart from binary garbage in a corrupt/hostile dentry.
int utf8_sequence_len(const unsigned char* p, std::size_t avail) {
    const unsigned char c = p[0];
    if (c < 0x80) return 1;
    auto cont = [](unsigned char b, unsigned char lo, unsigned char hi) {
        return b >= lo && b <= hi;
    };
    if (c >= 0xC2 && c <= 0xDF)
        return (avail >= 2 && cont(p[1], 0x80, 0xBF)) ? 2 : 0;
    if (c == 0xE0)
        return (avail >= 3 && cont(p[1], 0xA0, 0xBF) && cont(p[2], 0x80, 0xBF)) ? 3 : 0;
    if (c >= 0xE1 && c <= 0xEC)
        return (avail >= 3 && cont(p[1], 0x80, 0xBF) && cont(p[2], 0x80, 0xBF)) ? 3 : 0;
    if (c == 0xED)   // exclude surrogates: second byte capped at 0x9F
        return (avail >= 3 && cont(p[1], 0x80, 0x9F) && cont(p[2], 0x80, 0xBF)) ? 3 : 0;
    if (c >= 0xEE && c <= 0xEF)
        return (avail >= 3 && cont(p[1], 0x80, 0xBF) && cont(p[2], 0x80, 0xBF)) ? 3 : 0;
    if (c == 0xF0)
        return (avail >= 4 && cont(p[1], 0x90, 0xBF) && cont(p[2], 0x80, 0xBF) &&
                cont(p[3], 0x80, 0xBF)) ? 4 : 0;
    if (c >= 0xF1 && c <= 0xF3)
        return (avail >= 4 && cont(p[1], 0x80, 0xBF) && cont(p[2], 0x80, 0xBF) &&
                cont(p[3], 0x80, 0xBF)) ? 4 : 0;
    if (c == 0xF4)   // cap at U+10FFFF: second byte 0x80–0x8F
        return (avail >= 4 && cont(p[1], 0x80, 0x8F) && cont(p[2], 0x80, 0xBF) &&
                cont(p[3], 0x80, 0xBF)) ? 4 : 0;
    return 0;        // 0x80–0xBF lone, 0xC0/0xC1 overlong, 0xF5–0xFF out of range
}

PathTrust validate_path_component(std::string_view comp) {
    if (comp.empty()) return {false, "empty path component"};