1 // SPDX-License-Identifier: GPL-2.0+
2 /*
3 * EFI Unicode collation protocol
4 *
5 * Copyright (c) 2018 Heinrich Schuchardt <xypron.glpk@gmx.de>
6 */
7
8 #include <common.h>
9 #include <charset.h>
10 #include <cp1250.h>
11 #include <cp437.h>
12 #include <efi_loader.h>
13
14 /* Characters that may not be used in file names */
15 static const char illegal[] = "<>:\"/\\|?*";
16
17 /*
18 * EDK2 assumes codepage 1250 when creating FAT 8.3 file names.
19 * Linux defaults to codepage 437 for FAT 8.3 file names.
20 */
21 #if CONFIG_FAT_DEFAULT_CODEPAGE == 1250
22 /* Unicode code points for code page 1250 characters 0x80 - 0xff */
23 static const u16 codepage[] = CP1250;
24 #else
25 /* Unicode code points for code page 437 characters 0x80 - 0xff */
26 static const u16 codepage[] = CP437;
27 #endif
28
29 /* GUID of the EFI_UNICODE_COLLATION_PROTOCOL */
30 const efi_guid_t efi_guid_unicode_collation_protocol =
31 EFI_UNICODE_COLLATION_PROTOCOL2_GUID;
32
33 /**
34 * efi_stri_coll() - compare utf-16 strings case-insenitively
35 *
36 * @this: unicode collation protocol instance
37 * @s1: first string
38 * @s2: second string
39 *
40 * This function implements the StriColl() service of the
41 * EFI_UNICODE_COLLATION_PROTOCOL.
42 *
43 * See the Unified Extensible Firmware Interface (UEFI) specification for
44 * details.
45 *
46 * TODO:
47 * The implementation does not follow the Unicode collation algorithm.
48 * For ASCII characters it results in the same sort order as EDK2.
49 * We could use table UNICODE_CAPITALIZATION_TABLE for better results.
50 *
51 * Return: 0: s1 == s2, > 0: s1 > s2, < 0: s1 < s2
52 */
efi_stri_coll(struct efi_unicode_collation_protocol * this,u16 * s1,u16 * s2)53 static efi_intn_t EFIAPI efi_stri_coll(
54 struct efi_unicode_collation_protocol *this, u16 *s1, u16 *s2)
55 {
56 s32 c1, c2;
57 efi_intn_t ret = 0;
58
59 EFI_ENTRY("%p, %ls, %ls", this, s1, s2);
60 for (; *s1 | *s2; ++s1, ++s2) {
61 c1 = utf_to_upper(*s1);
62 c2 = utf_to_upper(*s2);
63 if (c1 < c2) {
64 ret = -1;
65 goto out;
66 } else if (c1 > c2) {
67 ret = 1;
68 goto out;
69 }
70 }
71 out:
72 EFI_EXIT(EFI_SUCCESS);
73 return ret;
74 }
75
76 /**
77 * metai_match() - compare utf-16 string with a pattern string case-insenitively
78 *
79 * @s: string to compare
80 * @p: pattern string
81 *
82 * The pattern string may use these:
83 * - * matches >= 0 characters
84 * - ? matches 1 character
85 * - [<char1><char2>...<charN>] match any character in the set
86 * - [<char1>-<char2>] matches any character in the range
87 *
88 * This function is called my efi_metai_match().
89 *
90 * For '*' pattern searches this function calls itself recursively.
91 * Performance-wise this is suboptimal, especially for multiple '*' wildcards.
92 * But it results in simple code.
93 *
94 * Return: true if the string is matched.
95 */
metai_match(const u16 * s,const u16 * p)96 static bool metai_match(const u16 *s, const u16 *p)
97 {
98 u16 first;
99
100 for (; *s && *p; ++s, ++p) {
101 switch (*p) {
102 case '*':
103 /* Match 0 or more characters */
104 ++p;
105 for (;; ++s) {
106 if (metai_match(s, p))
107 return true;
108 if (!*s)
109 return false;
110 }
111 case '?':
112 /* Match any one character */
113 break;
114 case '[':
115 /* Match any character in the set */
116 ++p;
117 first = *p;
118 if (first == ']')
119 /* Empty set */
120 return false;
121 ++p;
122 if (*p == '-') {
123 /* Range */
124 ++p;
125 if (*s < first || *s > *p)
126 return false;
127 ++p;
128 if (*p != ']')
129 return false;
130 } else {
131 /* Set */
132 bool hit = false;
133
134 if (*s == first)
135 hit = true;
136 for (; *p && *p != ']'; ++p) {
137 if (*p == *s)
138 hit = true;
139 }
140 if (!hit || *p != ']')
141 return false;
142 }
143 break;
144 default:
145 /* Match one character */
146 if (*p != *s)
147 return false;
148 }
149 }
150 if (!*p && !*s)
151 return true;
152 return false;
153 }
154
155 /**
156 * efi_metai_match() - compare utf-16 string with a pattern string
157 * case-insenitively
158 *
159 * @this: unicode collation protocol instance
160 * @s: string to compare
161 * @p: pattern string
162 *
163 * The pattern string may use these:
164 * - * matches >= 0 characters
165 * - ? matches 1 character
166 * - [<char1><char2>...<charN>] match any character in the set
167 * - [<char1>-<char2>] matches any character in the range
168 *
169 * This function implements the MetaMatch() service of the
170 * EFI_UNICODE_COLLATION_PROTOCOL.
171 *
172 * Return: true if the string is matched.
173 */
efi_metai_match(struct efi_unicode_collation_protocol * this,const u16 * string,const u16 * pattern)174 static bool EFIAPI efi_metai_match(struct efi_unicode_collation_protocol *this,
175 const u16 *string, const u16 *pattern)
176 {
177 bool ret;
178
179 EFI_ENTRY("%p, %ls, %ls", this, string, pattern);
180 ret = metai_match(string, pattern);
181 EFI_EXIT(EFI_SUCCESS);
182 return ret;
183 }
184
185 /**
186 * efi_str_lwr() - convert to lower case
187 *
188 * @this: unicode collation protocol instance
189 * @string: string to convert
190 * @p: pattern string
191 *
192 * The conversion is done in place. As long as upper and lower letters use the
193 * same number of words this does not pose a problem.
194 *
195 * This function implements the StrLwr() service of the
196 * EFI_UNICODE_COLLATION_PROTOCOL.
197 */
efi_str_lwr(struct efi_unicode_collation_protocol * this,u16 * string)198 static void EFIAPI efi_str_lwr(struct efi_unicode_collation_protocol *this,
199 u16 *string)
200 {
201 EFI_ENTRY("%p, %ls", this, string);
202 for (; *string; ++string)
203 *string = utf_to_lower(*string);
204 EFI_EXIT(EFI_SUCCESS);
205 }
206
207 /**
208 * efi_str_upr() - convert to upper case
209 *
210 * @this: unicode collation protocol instance
211 * @string: string to convert
212 * @p: pattern string
213 *
214 * The conversion is done in place. As long as upper and lower letters use the
215 * same number of words this does not pose a problem.
216 *
217 * This function implements the StrUpr() service of the
218 * EFI_UNICODE_COLLATION_PROTOCOL.
219 */
efi_str_upr(struct efi_unicode_collation_protocol * this,u16 * string)220 static void EFIAPI efi_str_upr(struct efi_unicode_collation_protocol *this,
221 u16 *string)
222 {
223 EFI_ENTRY("%p, %ls", this, string);
224 for (; *string; ++string)
225 *string = utf_to_upper(*string);
226 EFI_EXIT(EFI_SUCCESS);
227 }
228
229 /**
230 * efi_fat_to_str() - convert an 8.3 file name from an OEM codepage to Unicode
231 *
232 * @this: unicode collation protocol instance
233 * @fat_size: size of the string to convert
234 * @fat: string to convert
235 * @string: converted string
236 *
237 * This function implements the FatToStr() service of the
238 * EFI_UNICODE_COLLATION_PROTOCOL.
239 */
efi_fat_to_str(struct efi_unicode_collation_protocol * this,efi_uintn_t fat_size,char * fat,u16 * string)240 static void EFIAPI efi_fat_to_str(struct efi_unicode_collation_protocol *this,
241 efi_uintn_t fat_size, char *fat, u16 *string)
242 {
243 efi_uintn_t i;
244 u16 c;
245
246 EFI_ENTRY("%p, %zu, %s, %p", this, fat_size, fat, string);
247 for (i = 0; i < fat_size; ++i) {
248 c = (unsigned char)fat[i];
249 if (c > 0x80)
250 c = codepage[i - 0x80];
251 string[i] = c;
252 if (!c)
253 break;
254 }
255 string[i] = 0;
256 EFI_EXIT(EFI_SUCCESS);
257 }
258
259 /**
260 * efi_fat_to_str() - convert a utf-16 string to legal characters for a FAT
261 * file name in an OEM code page
262 *
263 * @this: unicode collation protocol instance
264 * @string: Unicode string to convert
265 * @fat_size: size of the target buffer
266 * @fat: converted string
267 *
268 * This function implements the StrToFat() service of the
269 * EFI_UNICODE_COLLATION_PROTOCOL.
270 *
271 * Return: true if an illegal character was substituted by '_'.
272 */
efi_str_to_fat(struct efi_unicode_collation_protocol * this,const u16 * string,efi_uintn_t fat_size,char * fat)273 static bool EFIAPI efi_str_to_fat(struct efi_unicode_collation_protocol *this,
274 const u16 *string, efi_uintn_t fat_size,
275 char *fat)
276 {
277 efi_uintn_t i;
278 s32 c;
279 bool ret = false;
280
281 EFI_ENTRY("%p, %ls, %zu, %p", this, string, fat_size, fat);
282 for (i = 0; i < fat_size;) {
283 c = utf16_get(&string);
284 switch (c) {
285 /* Ignore period and space */
286 case '.':
287 case ' ':
288 continue;
289 case 0:
290 break;
291 }
292 c = utf_to_upper(c);
293 if (c >= 0x80) {
294 int j;
295
296 /* Look for codepage translation */
297 for (j = 0; j < 0x80; ++j) {
298 if (c == codepage[j]) {
299 c = j + 0x80;
300 break;
301 }
302 }
303 if (j >= 0x80) {
304 c = '_';
305 ret = true;
306 }
307 } else if (c && (c < 0x20 || strchr(illegal, c))) {
308 c = '_';
309 ret = true;
310 }
311
312 fat[i] = c;
313 if (!c)
314 break;
315 ++i;
316 }
317 EFI_EXIT(EFI_SUCCESS);
318 return ret;
319 }
320
321 const struct efi_unicode_collation_protocol efi_unicode_collation_protocol = {
322 .stri_coll = efi_stri_coll,
323 .metai_match = efi_metai_match,
324 .str_lwr = efi_str_lwr,
325 .str_upr = efi_str_upr,
326 .fat_to_str = efi_fat_to_str,
327 .str_to_fat = efi_str_to_fat,
328 .supported_languages = "en",
329 };
330