lib: add module support to glob tests
authorGeert Uytterhoeven <geert@linux-m68k.org>
Fri, 24 Feb 2017 23:00:52 +0000 (15:00 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 25 Feb 2017 01:46:57 +0000 (17:46 -0800)
Extract the glob test code into its own source file, to allow to compile
it either to a loadable module, or builtin into the kernel.

Link: http://lkml.kernel.org/r/1483470276-10517-2-git-send-email-geert@linux-m68k.org
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
lib/Kconfig
lib/Makefile
lib/glob.c
lib/globtest.c [new file with mode: 0644]

index ffd7635efda609bcd42a685e875fab79d8352621..fe7e8e175db8f7b4c3e72f62c6dffeb01d0ecaee 100644 (file)
@@ -431,8 +431,7 @@ config GLOB
          depends on this.
 
 config GLOB_SELFTEST
-       bool "glob self-test on init"
-       default n
+       tristate "glob self-test on init"
        depends on GLOB
        help
          This option enables a simple self-test of the glob_match
index bc9be67b5f8b9caa95b06941cd0293b34ea65cca..d6d53b70f58dcc5f6500f75c91c6f66d709d5e41 100644 (file)
@@ -162,6 +162,7 @@ obj-$(CONFIG_CORDIC) += cordic.o
 obj-$(CONFIG_DQL) += dynamic_queue_limits.o
 
 obj-$(CONFIG_GLOB) += glob.o
+obj-$(CONFIG_GLOB_SELFTEST) += globtest.o
 
 obj-$(CONFIG_MPILIB) += mpi/
 obj-$(CONFIG_SIGNATURE) += digsig.o
index 500fc80d23e107651d8075603a343e4c13629a64..0ba3ea86b5466c102d588327dad3e7e8df092f5d 100644 (file)
@@ -121,167 +121,3 @@ backtrack:
        }
 }
 EXPORT_SYMBOL(glob_match);
-
-
-#ifdef CONFIG_GLOB_SELFTEST
-
-#include <linux/printk.h>
-#include <linux/moduleparam.h>
-
-/* Boot with "glob.verbose=1" to show successful tests, too */
-static bool verbose = false;
-module_param(verbose, bool, 0);
-
-struct glob_test {
-       char const *pat, *str;
-       bool expected;
-};
-
-static bool __pure __init test(char const *pat, char const *str, bool expected)
-{
-       bool match = glob_match(pat, str);
-       bool success = match == expected;
-
-       /* Can't get string literals into a particular section, so... */
-       static char const msg_error[] __initconst =
-               KERN_ERR "glob: \"%s\" vs. \"%s\": %s *** ERROR ***\n";
-       static char const msg_ok[] __initconst =
-               KERN_DEBUG "glob: \"%s\" vs. \"%s\": %s OK\n";
-       static char const mismatch[] __initconst = "mismatch";
-       char const *message;
-
-       if (!success)
-               message = msg_error;
-       else if (verbose)
-               message = msg_ok;
-       else
-               return success;
-
-       printk(message, pat, str, mismatch + 3*match);
-       return success;
-}
-
-/*
- * The tests are all jammed together in one array to make it simpler
- * to place that array in the .init.rodata section.  The obvious
- * "array of structures containing char *" has no way to force the
- * pointed-to strings to be in a particular section.
- *
- * Anyway, a test consists of:
- * 1. Expected glob_match result: '1' or '0'.
- * 2. Pattern to match: null-terminated string
- * 3. String to match against: null-terminated string
- *
- * The list of tests is terminated with a final '\0' instead of
- * a glob_match result character.
- */
-static char const glob_tests[] __initconst =
-       /* Some basic tests */
-       "1" "a\0" "a\0"
-       "0" "a\0" "b\0"
-       "0" "a\0" "aa\0"
-       "0" "a\0" "\0"
-       "1" "\0" "\0"
-       "0" "\0" "a\0"
-       /* Simple character class tests */
-       "1" "[a]\0" "a\0"
-       "0" "[a]\0" "b\0"
-       "0" "[!a]\0" "a\0"
-       "1" "[!a]\0" "b\0"
-       "1" "[ab]\0" "a\0"
-       "1" "[ab]\0" "b\0"
-       "0" "[ab]\0" "c\0"
-       "1" "[!ab]\0" "c\0"
-       "1" "[a-c]\0" "b\0"
-       "0" "[a-c]\0" "d\0"
-       /* Corner cases in character class parsing */
-       "1" "[a-c-e-g]\0" "-\0"
-       "0" "[a-c-e-g]\0" "d\0"
-       "1" "[a-c-e-g]\0" "f\0"
-       "1" "[]a-ceg-ik[]\0" "a\0"
-       "1" "[]a-ceg-ik[]\0" "]\0"
-       "1" "[]a-ceg-ik[]\0" "[\0"
-       "1" "[]a-ceg-ik[]\0" "h\0"
-       "0" "[]a-ceg-ik[]\0" "f\0"
-       "0" "[!]a-ceg-ik[]\0" "h\0"
-       "0" "[!]a-ceg-ik[]\0" "]\0"
-       "1" "[!]a-ceg-ik[]\0" "f\0"
-       /* Simple wild cards */
-       "1" "?\0" "a\0"
-       "0" "?\0" "aa\0"
-       "0" "??\0" "a\0"
-       "1" "?x?\0" "axb\0"
-       "0" "?x?\0" "abx\0"
-       "0" "?x?\0" "xab\0"
-       /* Asterisk wild cards (backtracking) */
-       "0" "*??\0" "a\0"
-       "1" "*??\0" "ab\0"
-       "1" "*??\0" "abc\0"
-       "1" "*??\0" "abcd\0"
-       "0" "??*\0" "a\0"
-       "1" "??*\0" "ab\0"
-       "1" "??*\0" "abc\0"
-       "1" "??*\0" "abcd\0"
-       "0" "?*?\0" "a\0"
-       "1" "?*?\0" "ab\0"
-       "1" "?*?\0" "abc\0"
-       "1" "?*?\0" "abcd\0"
-       "1" "*b\0" "b\0"
-       "1" "*b\0" "ab\0"
-       "0" "*b\0" "ba\0"
-       "1" "*b\0" "bb\0"
-       "1" "*b\0" "abb\0"
-       "1" "*b\0" "bab\0"
-       "1" "*bc\0" "abbc\0"
-       "1" "*bc\0" "bc\0"
-       "1" "*bc\0" "bbc\0"
-       "1" "*bc\0" "bcbc\0"
-       /* Multiple asterisks (complex backtracking) */
-       "1" "*ac*\0" "abacadaeafag\0"
-       "1" "*ac*ae*ag*\0" "abacadaeafag\0"
-       "1" "*a*b*[bc]*[ef]*g*\0" "abacadaeafag\0"
-       "0" "*a*b*[ef]*[cd]*g*\0" "abacadaeafag\0"
-       "1" "*abcd*\0" "abcabcabcabcdefg\0"
-       "1" "*ab*cd*\0" "abcabcabcabcdefg\0"
-       "1" "*abcd*abcdef*\0" "abcabcdabcdeabcdefg\0"
-       "0" "*abcd*\0" "abcabcabcabcefg\0"
-       "0" "*ab*cd*\0" "abcabcabcabcefg\0";
-
-static int __init glob_init(void)
-{
-       unsigned successes = 0;
-       unsigned n = 0;
-       char const *p = glob_tests;
-       static char const message[] __initconst =
-               KERN_INFO "glob: %u self-tests passed, %u failed\n";
-
-       /*
-        * Tests are jammed together in a string.  The first byte is '1'
-        * or '0' to indicate the expected outcome, or '\0' to indicate the
-        * end of the tests.  Then come two null-terminated strings: the
-        * pattern and the string to match it against.
-        */
-       while (*p) {
-               bool expected = *p++ & 1;
-               char const *pat = p;
-
-               p += strlen(p) + 1;
-               successes += test(pat, p, expected);
-               p += strlen(p) + 1;
-               n++;
-       }
-
-       n -= successes;
-       printk(message, successes, n);
-
-       /* What's the errno for "kernel bug detected"?  Guess... */
-       return n ? -ECANCELED : 0;
-}
-
-/* We need a dummy exit function to allow unload */
-static void __exit glob_fini(void) { }
-
-module_init(glob_init);
-module_exit(glob_fini);
-
-#endif /* CONFIG_GLOB_SELFTEST */
diff --git a/lib/globtest.c b/lib/globtest.c
new file mode 100644 (file)
index 0000000..d8e97d4
--- /dev/null
@@ -0,0 +1,167 @@
+/*
+ * Extracted fronm glob.c
+ */
+
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/glob.h>
+#include <linux/printk.h>
+
+/* Boot with "glob.verbose=1" to show successful tests, too */
+static bool verbose = false;
+module_param(verbose, bool, 0);
+
+struct glob_test {
+       char const *pat, *str;
+       bool expected;
+};
+
+static bool __pure __init test(char const *pat, char const *str, bool expected)
+{
+       bool match = glob_match(pat, str);
+       bool success = match == expected;
+
+       /* Can't get string literals into a particular section, so... */
+       static char const msg_error[] __initconst =
+               KERN_ERR "glob: \"%s\" vs. \"%s\": %s *** ERROR ***\n";
+       static char const msg_ok[] __initconst =
+               KERN_DEBUG "glob: \"%s\" vs. \"%s\": %s OK\n";
+       static char const mismatch[] __initconst = "mismatch";
+       char const *message;
+
+       if (!success)
+               message = msg_error;
+       else if (verbose)
+               message = msg_ok;
+       else
+               return success;
+
+       printk(message, pat, str, mismatch + 3*match);
+       return success;
+}
+
+/*
+ * The tests are all jammed together in one array to make it simpler
+ * to place that array in the .init.rodata section.  The obvious
+ * "array of structures containing char *" has no way to force the
+ * pointed-to strings to be in a particular section.
+ *
+ * Anyway, a test consists of:
+ * 1. Expected glob_match result: '1' or '0'.
+ * 2. Pattern to match: null-terminated string
+ * 3. String to match against: null-terminated string
+ *
+ * The list of tests is terminated with a final '\0' instead of
+ * a glob_match result character.
+ */
+static char const glob_tests[] __initconst =
+       /* Some basic tests */
+       "1" "a\0" "a\0"
+       "0" "a\0" "b\0"
+       "0" "a\0" "aa\0"
+       "0" "a\0" "\0"
+       "1" "\0" "\0"
+       "0" "\0" "a\0"
+       /* Simple character class tests */
+       "1" "[a]\0" "a\0"
+       "0" "[a]\0" "b\0"
+       "0" "[!a]\0" "a\0"
+       "1" "[!a]\0" "b\0"
+       "1" "[ab]\0" "a\0"
+       "1" "[ab]\0" "b\0"
+       "0" "[ab]\0" "c\0"
+       "1" "[!ab]\0" "c\0"
+       "1" "[a-c]\0" "b\0"
+       "0" "[a-c]\0" "d\0"
+       /* Corner cases in character class parsing */
+       "1" "[a-c-e-g]\0" "-\0"
+       "0" "[a-c-e-g]\0" "d\0"
+       "1" "[a-c-e-g]\0" "f\0"
+       "1" "[]a-ceg-ik[]\0" "a\0"
+       "1" "[]a-ceg-ik[]\0" "]\0"
+       "1" "[]a-ceg-ik[]\0" "[\0"
+       "1" "[]a-ceg-ik[]\0" "h\0"
+       "0" "[]a-ceg-ik[]\0" "f\0"
+       "0" "[!]a-ceg-ik[]\0" "h\0"
+       "0" "[!]a-ceg-ik[]\0" "]\0"
+       "1" "[!]a-ceg-ik[]\0" "f\0"
+       /* Simple wild cards */
+       "1" "?\0" "a\0"
+       "0" "?\0" "aa\0"
+       "0" "??\0" "a\0"
+       "1" "?x?\0" "axb\0"
+       "0" "?x?\0" "abx\0"
+       "0" "?x?\0" "xab\0"
+       /* Asterisk wild cards (backtracking) */
+       "0" "*??\0" "a\0"
+       "1" "*??\0" "ab\0"
+       "1" "*??\0" "abc\0"
+       "1" "*??\0" "abcd\0"
+       "0" "??*\0" "a\0"
+       "1" "??*\0" "ab\0"
+       "1" "??*\0" "abc\0"
+       "1" "??*\0" "abcd\0"
+       "0" "?*?\0" "a\0"
+       "1" "?*?\0" "ab\0"
+       "1" "?*?\0" "abc\0"
+       "1" "?*?\0" "abcd\0"
+       "1" "*b\0" "b\0"
+       "1" "*b\0" "ab\0"
+       "0" "*b\0" "ba\0"
+       "1" "*b\0" "bb\0"
+       "1" "*b\0" "abb\0"
+       "1" "*b\0" "bab\0"
+       "1" "*bc\0" "abbc\0"
+       "1" "*bc\0" "bc\0"
+       "1" "*bc\0" "bbc\0"
+       "1" "*bc\0" "bcbc\0"
+       /* Multiple asterisks (complex backtracking) */
+       "1" "*ac*\0" "abacadaeafag\0"
+       "1" "*ac*ae*ag*\0" "abacadaeafag\0"
+       "1" "*a*b*[bc]*[ef]*g*\0" "abacadaeafag\0"
+       "0" "*a*b*[ef]*[cd]*g*\0" "abacadaeafag\0"
+       "1" "*abcd*\0" "abcabcabcabcdefg\0"
+       "1" "*ab*cd*\0" "abcabcabcabcdefg\0"
+       "1" "*abcd*abcdef*\0" "abcabcdabcdeabcdefg\0"
+       "0" "*abcd*\0" "abcabcabcabcefg\0"
+       "0" "*ab*cd*\0" "abcabcabcabcefg\0";
+
+static int __init glob_init(void)
+{
+       unsigned successes = 0;
+       unsigned n = 0;
+       char const *p = glob_tests;
+       static char const message[] __initconst =
+               KERN_INFO "glob: %u self-tests passed, %u failed\n";
+
+       /*
+        * Tests are jammed together in a string.  The first byte is '1'
+        * or '0' to indicate the expected outcome, or '\0' to indicate the
+        * end of the tests.  Then come two null-terminated strings: the
+        * pattern and the string to match it against.
+        */
+       while (*p) {
+               bool expected = *p++ & 1;
+               char const *pat = p;
+
+               p += strlen(p) + 1;
+               successes += test(pat, p, expected);
+               p += strlen(p) + 1;
+               n++;
+       }
+
+       n -= successes;
+       printk(message, successes, n);
+
+       /* What's the errno for "kernel bug detected"?  Guess... */
+       return n ? -ECANCELED : 0;
+}
+
+/* We need a dummy exit function to allow unload */
+static void __exit glob_fini(void) { }
+
+module_init(glob_init);
+module_exit(glob_fini);
+
+MODULE_DESCRIPTION("glob(7) matching tests");
+MODULE_LICENSE("Dual MIT/GPL");