1 // SPDX-License-Identifier: (GPL-2.0 OR MIT)
2 #include <linux/module.h>
3 #include <linux/glob.h>
4 #include <linux/export.h>
5
6 /*
7 * The only reason this code can be compiled as a module is because the
8 * ATA code that depends on it can be as well. In practice, they're
9 * both usually compiled in and the module overhead goes away.
10 */
11 MODULE_DESCRIPTION("glob(7) matching");
12 MODULE_LICENSE("Dual MIT/GPL");
13
14 /**
15 * glob_match - Shell-style pattern matching, like !fnmatch(pat, str, 0)
16 * @pat: Shell-style pattern to match, e.g. "*.[ch]".
17 * @str: String to match. The pattern must match the entire string.
18 *
19 * Perform shell-style glob matching, returning true (1) if the match
20 * succeeds, or false (0) if it fails. Equivalent to !fnmatch(@pat, @str, 0).
21 *
22 * Pattern metacharacters are ?, *, [ and \.
23 * (And, inside character classes, !, - and ].)
24 *
25 * This is a small and simple implementation intended for device denylists
26 * where a string is matched against a number of patterns. Thus, it
27 * does not preprocess the patterns. It is non-recursive, and run-time
28 * is at most quadratic: strlen(@str)*strlen(@pat).
29 *
30 * An example of the worst case is glob_match("*aaaaa", "aaaaaaaaaa");
31 * it takes 6 passes over the pattern before matching the string.
32 *
33 * Like !fnmatch(@pat, @str, 0) and unlike the shell, this does NOT
34 * treat / or leading . specially; it isn't actually used for pathnames.
35 *
36 * Note that according to glob(7) (and unlike bash), character classes
37 * are complemented by a leading !; this does not support the regex-style
38 * [^a-z] syntax.
39 *
40 * An opening bracket without a matching close is matched literally.
41 */
glob_match(char const * pat,char const * str)42 bool __pure glob_match(char const *pat, char const *str)
43 {
44 /*
45 * Backtrack to previous * on mismatch and retry starting one
46 * character later in the string. Because * matches all characters
47 * (no exception for /), it can be easily proved that there's
48 * never a need to backtrack multiple levels.
49 */
50 char const *back_pat = NULL, *back_str = NULL;
51
52 /*
53 * Loop over each token (character or class) in pat, matching
54 * it against the remaining unmatched tail of str. Return false
55 * on mismatch, or true after matching the trailing nul bytes.
56 */
57 for (;;) {
58 unsigned char c = *str++;
59 unsigned char d = *pat++;
60
61 switch (d) {
62 case '?': /* Wildcard: anything but nul */
63 if (c == '\0')
64 return false;
65 break;
66 case '*': /* Any-length wildcard */
67 if (*pat == '\0') /* Optimize trailing * case */
68 return true;
69 back_pat = pat;
70 back_str = --str; /* Allow zero-length match */
71 break;
72 case '[': { /* Character class */
73 if (c == '\0') /* No possible match */
74 return false;
75 bool match = false, inverted = (*pat == '!');
76 char const *class = inverted ? pat + 1 : pat;
77 unsigned char a = *class++;
78
79 /*
80 * Iterate over each span in the character class.
81 * A span is either a single character a, or a
82 * range a-b. The first span may begin with ']'.
83 */
84 do {
85 unsigned char b = a;
86
87 if (a == '\0') /* Malformed */
88 goto literal;
89
90 if (class[0] == '-' && class[1] != ']') {
91 b = class[1];
92
93 if (b == '\0')
94 goto literal;
95
96 class += 2;
97 /* Any special action if a > b? */
98 }
99 if (a <= c && c <= b)
100 match = true;
101 } while ((a = *class++) != ']');
102
103 if (match == inverted)
104 goto backtrack;
105 pat = class;
106 }
107 break;
108 case '\\':
109 d = *pat++;
110 fallthrough;
111 default: /* Literal character */
112 literal:
113 if (c == d) {
114 if (d == '\0')
115 return true;
116 break;
117 }
118 backtrack:
119 if (c == '\0' || !back_pat)
120 return false; /* No point continuing */
121 /* Try again from last *, one character later in str. */
122 pat = back_pat;
123 str = ++back_str;
124 break;
125 }
126 }
127 }
128 EXPORT_SYMBOL(glob_match);
129