| Li Jun | e787da4 | 2021-01-25 21:43:47 +0800 | [diff] [blame] | 1 | #ifndef _LINUX_UTF_H |
| 2 | #define _LINUX_UTF_H |
| 3 | |
| 4 | #include <asm/unaligned.h> |
| 5 | |
| 6 | static inline int utf8_to_utf16le(const char *s, __le16 *cp, unsigned len) |
| 7 | { |
| 8 | int count = 0; |
| 9 | u8 c; |
| 10 | u16 uchar; |
| 11 | |
| 12 | /* |
| 13 | * this insists on correct encodings, though not minimal ones. |
| 14 | * BUT it currently rejects legit 4-byte UTF-8 code points, |
| 15 | * which need surrogate pairs. (Unicode 3.1 can use them.) |
| 16 | */ |
| 17 | while (len != 0 && (c = (u8) *s++) != 0) { |
| 18 | if ((c & 0x80)) { |
| 19 | /* |
| 20 | * 2-byte sequence: |
| 21 | * 00000yyyyyxxxxxx = 110yyyyy 10xxxxxx |
| 22 | */ |
| 23 | if ((c & 0xe0) == 0xc0) { |
| 24 | uchar = (c & 0x1f) << 6; |
| 25 | |
| 26 | c = (u8) *s++; |
| 27 | if ((c & 0xc0) != 0x80) |
| 28 | goto fail; |
| 29 | c &= 0x3f; |
| 30 | uchar |= c; |
| 31 | |
| 32 | /* |
| 33 | * 3-byte sequence (most CJKV characters): |
| 34 | * zzzzyyyyyyxxxxxx = 1110zzzz 10yyyyyy 10xxxxxx |
| 35 | */ |
| 36 | } else if ((c & 0xf0) == 0xe0) { |
| 37 | uchar = (c & 0x0f) << 12; |
| 38 | |
| 39 | c = (u8) *s++; |
| 40 | if ((c & 0xc0) != 0x80) |
| 41 | goto fail; |
| 42 | c &= 0x3f; |
| 43 | uchar |= c << 6; |
| 44 | |
| 45 | c = (u8) *s++; |
| 46 | if ((c & 0xc0) != 0x80) |
| 47 | goto fail; |
| 48 | c &= 0x3f; |
| 49 | uchar |= c; |
| 50 | |
| 51 | /* no bogus surrogates */ |
| 52 | if (0xd800 <= uchar && uchar <= 0xdfff) |
| 53 | goto fail; |
| 54 | |
| 55 | /* |
| 56 | * 4-byte sequence (surrogate pairs, currently rare): |
| 57 | * 11101110wwwwzzzzyy + 110111yyyyxxxxxx |
| 58 | * = 11110uuu 10uuzzzz 10yyyyyy 10xxxxxx |
| 59 | * (uuuuu = wwww + 1) |
| 60 | * FIXME accept the surrogate code points (only) |
| 61 | */ |
| 62 | } else |
| 63 | goto fail; |
| 64 | } else |
| 65 | uchar = c; |
| 66 | put_unaligned_le16(uchar, cp++); |
| 67 | count++; |
| 68 | len--; |
| 69 | } |
| 70 | return count; |
| 71 | fail: |
| 72 | return -1; |
| 73 | } |
| 74 | |
| 75 | #endif /* _LINUX_UTF_H */ |