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 FAT 8.3 file names */
15 static const char illegal[] = "+,<=>:;\"/\\|?*[]\x7f";
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 = codepage_437;
27 #endif
28 
29 /* GUID of the EFI_UNICODE_COLLATION_PROTOCOL2 */
30 const efi_guid_t efi_guid_unicode_collation_protocol2 =
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_PROTOCOL2.
42  *
43  * See the Unified Extensible Firmware Interface (UEFI) specification for
44  * details.
45  *
46  * Return:	0: s1 == s2, > 0: s1 > s2, < 0: s1 < s2
47  */
efi_stri_coll(struct efi_unicode_collation_protocol * this,u16 * s1,u16 * s2)48 static efi_intn_t EFIAPI efi_stri_coll(
49 		struct efi_unicode_collation_protocol *this, u16 *s1, u16 *s2)
50 {
51 	s32 c1, c2;
52 	efi_intn_t ret = 0;
53 
54 	EFI_ENTRY("%p, %ls, %ls", this, s1, s2);
55 	for (; *s1 | *s2; ++s1, ++s2) {
56 		c1 = utf_to_upper(*s1);
57 		c2 = utf_to_upper(*s2);
58 		if (c1 < c2) {
59 			ret = -1;
60 			goto out;
61 		} else if (c1 > c2) {
62 			ret = 1;
63 			goto out;
64 		}
65 	}
66 out:
67 	EFI_EXIT(EFI_SUCCESS);
68 	return ret;
69 }
70 
71 /**
72  * next_lower() - get next codepoint converted to lower case
73  *
74  * @string:	pointer to u16 string, on return advanced by one codepoint
75  * Return:	first codepoint of string converted to lower case
76  */
next_lower(const u16 ** string)77 static s32 next_lower(const u16 **string)
78 {
79 	return utf_to_lower(utf16_get(string));
80 }
81 
82 /**
83  * metai_match() - compare utf-16 string with a pattern string case-insenitively
84  *
85  * @string:	string to compare
86  * @pattern:	pattern string
87  *
88  * The pattern string may use these:
89  *	- * matches >= 0 characters
90  *	- ? matches 1 character
91  *	- [<char1><char2>...<charN>] match any character in the set
92  *	- [<char1>-<char2>] matches any character in the range
93  *
94  * This function is called my efi_metai_match().
95  *
96  * For '*' pattern searches this function calls itself recursively.
97  * Performance-wise this is suboptimal, especially for multiple '*' wildcards.
98  * But it results in simple code.
99  *
100  * Return:	true if the string is matched.
101  */
metai_match(const u16 * string,const u16 * pattern)102 static bool metai_match(const u16 *string, const u16 *pattern)
103 {
104 	s32 first, s, p;
105 
106 	for (; *string && *pattern;) {
107 		const u16 *string_old = string;
108 
109 		s = next_lower(&string);
110 		p = next_lower(&pattern);
111 
112 		switch (p) {
113 		case '*':
114 			/* Match 0 or more characters */
115 			for (;; s = next_lower(&string)) {
116 				if (metai_match(string_old, pattern))
117 					return true;
118 				if (!s)
119 					return false;
120 				string_old = string;
121 			}
122 		case '?':
123 			/* Match any one character */
124 			break;
125 		case '[':
126 			/* Match any character in the set */
127 			p = next_lower(&pattern);
128 			first = p;
129 			if (first == ']')
130 				/* Empty set */
131 				return false;
132 			p = next_lower(&pattern);
133 			if (p == '-') {
134 				/* Range */
135 				p = next_lower(&pattern);
136 				if (s < first || s > p)
137 					return false;
138 				p = next_lower(&pattern);
139 				if (p != ']')
140 					return false;
141 			} else {
142 				/* Set */
143 				bool hit = false;
144 
145 				if (s == first)
146 					hit = true;
147 				for (; p && p != ']';
148 				     p = next_lower(&pattern)) {
149 					if (p == s)
150 						hit = true;
151 				}
152 				if (!hit || p != ']')
153 					return false;
154 			}
155 			break;
156 		default:
157 			/* Match one character */
158 			if (p != s)
159 				return false;
160 		}
161 	}
162 	if (!*pattern && !*string)
163 		return true;
164 	return false;
165 }
166 
167 /**
168  * efi_metai_match() - compare utf-16 string with a pattern string
169  *		       case-insenitively
170  *
171  * @this:	unicode collation protocol instance
172  * @string:	string to compare
173  * @pattern:	pattern string
174  *
175  * The pattern string may use these:
176  *	- * matches >= 0 characters
177  *	- ? matches 1 character
178  *	- [<char1><char2>...<charN>] match any character in the set
179  *	- [<char1>-<char2>] matches any character in the range
180  *
181  * This function implements the MetaMatch() service of the
182  * EFI_UNICODE_COLLATION_PROTOCOL2.
183  *
184  * Return:	true if the string is matched.
185  */
efi_metai_match(struct efi_unicode_collation_protocol * this,const u16 * string,const u16 * pattern)186 static bool EFIAPI efi_metai_match(struct efi_unicode_collation_protocol *this,
187 				   const u16 *string, const u16 *pattern)
188 {
189 	bool ret;
190 
191 	EFI_ENTRY("%p, %ls, %ls", this, string, pattern);
192 	ret =  metai_match(string, pattern);
193 	EFI_EXIT(EFI_SUCCESS);
194 	return ret;
195 }
196 
197 /**
198  * efi_str_lwr() - convert to lower case
199  *
200  * @this:	unicode collation protocol instance
201  * @string:	string to convert
202  *
203  * The conversion is done in place. As long as upper and lower letters use the
204  * same number of words this does not pose a problem.
205  *
206  * This function implements the StrLwr() service of the
207  * EFI_UNICODE_COLLATION_PROTOCOL2.
208  */
efi_str_lwr(struct efi_unicode_collation_protocol * this,u16 * string)209 static void EFIAPI efi_str_lwr(struct efi_unicode_collation_protocol *this,
210 			       u16 *string)
211 {
212 	EFI_ENTRY("%p, %ls", this, string);
213 	for (; *string; ++string)
214 		*string = utf_to_lower(*string);
215 	EFI_EXIT(EFI_SUCCESS);
216 }
217 
218 /**
219  * efi_str_upr() - convert to upper case
220  *
221  * @this:	unicode collation protocol instance
222  * @string:	string to convert
223  *
224  * The conversion is done in place. As long as upper and lower letters use the
225  * same number of words this does not pose a problem.
226  *
227  * This function implements the StrUpr() service of the
228  * EFI_UNICODE_COLLATION_PROTOCOL2.
229  */
efi_str_upr(struct efi_unicode_collation_protocol * this,u16 * string)230 static void EFIAPI efi_str_upr(struct efi_unicode_collation_protocol *this,
231 			       u16 *string)
232 {
233 	EFI_ENTRY("%p, %ls", this, string);
234 	for (; *string; ++string)
235 		*string = utf_to_upper(*string);
236 	EFI_EXIT(EFI_SUCCESS);
237 }
238 
239 /**
240  * efi_fat_to_str() - convert an 8.3 file name from an OEM codepage to Unicode
241  *
242  * @this:	unicode collation protocol instance
243  * @fat_size:	size of the string to convert
244  * @fat:	string to convert
245  * @string:	converted string
246  *
247  * This function implements the FatToStr() service of the
248  * EFI_UNICODE_COLLATION_PROTOCOL2.
249  */
efi_fat_to_str(struct efi_unicode_collation_protocol * this,efi_uintn_t fat_size,char * fat,u16 * string)250 static void EFIAPI efi_fat_to_str(struct efi_unicode_collation_protocol *this,
251 				  efi_uintn_t fat_size, char *fat, u16 *string)
252 {
253 	efi_uintn_t i;
254 	u16 c;
255 
256 	EFI_ENTRY("%p, %zu, %s, %p", this, fat_size, fat, string);
257 	for (i = 0; i < fat_size; ++i) {
258 		c = (unsigned char)fat[i];
259 		if (c > 0x80)
260 			c = codepage[i - 0x80];
261 		string[i] = c;
262 		if (!c)
263 			break;
264 	}
265 	string[i] = 0;
266 	EFI_EXIT(EFI_SUCCESS);
267 }
268 
269 /**
270  * efi_fat_to_str() - convert a utf-16 string to legal characters for a FAT
271  *                    file name in an OEM code page
272  *
273  * @this:	unicode collation protocol instance
274  * @string:	Unicode string to convert
275  * @fat_size:	size of the target buffer
276  * @fat:	converted string
277  *
278  * This function implements the StrToFat() service of the
279  * EFI_UNICODE_COLLATION_PROTOCOL2.
280  *
281  * Return:	true if an illegal character was substituted by '_'.
282  */
efi_str_to_fat(struct efi_unicode_collation_protocol * this,const u16 * string,efi_uintn_t fat_size,char * fat)283 static bool EFIAPI efi_str_to_fat(struct efi_unicode_collation_protocol *this,
284 				  const u16 *string, efi_uintn_t fat_size,
285 				  char *fat)
286 {
287 	efi_uintn_t i;
288 	s32 c;
289 	bool ret = false;
290 
291 	EFI_ENTRY("%p, %ls, %zu, %p", this, string, fat_size, fat);
292 	for (i = 0; i < fat_size;) {
293 		c = utf16_get(&string);
294 		switch (c) {
295 		/* Ignore period and space */
296 		case '.':
297 		case ' ':
298 			continue;
299 		case 0:
300 			break;
301 		}
302 		c = utf_to_upper(c);
303 		if (utf_to_cp(&c, codepage) ||
304 		    (c && (c < 0x20 || strchr(illegal, c)))) {
305 			ret = true;
306 			c = '_';
307 		}
308 
309 		fat[i] = c;
310 		if (!c)
311 			break;
312 		++i;
313 	}
314 	EFI_EXIT(EFI_SUCCESS);
315 	return ret;
316 }
317 
318 const struct efi_unicode_collation_protocol efi_unicode_collation_protocol2 = {
319 	.stri_coll = efi_stri_coll,
320 	.metai_match = efi_metai_match,
321 	.str_lwr = efi_str_lwr,
322 	.str_upr = efi_str_upr,
323 	.fat_to_str = efi_fat_to_str,
324 	.str_to_fat = efi_str_to_fat,
325 	.supported_languages = "en",
326 };
327