summaryrefslogtreecommitdiff
path: root/lib
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2016-05-28 16:15:25 -0700
committerLinus Torvalds <torvalds@linux-foundation.org>2016-05-28 16:15:25 -0700
commit7e0fb73c52c4037b4d5ef9ff56c7296a3151bd92 (patch)
tree9ab023505d388563d937b3c3ac26ef3c2045dba2 /lib
parent4e8440b3b6b801953b2e53c55491cf98fc8f6c01 (diff)
parent4684fe95300c071983f77653e354c040fe80a265 (diff)
downloadlwn-7e0fb73c52c4037b4d5ef9ff56c7296a3151bd92.tar.gz
lwn-7e0fb73c52c4037b4d5ef9ff56c7296a3151bd92.zip
Merge branch 'hash' of git://ftp.sciencehorizons.net/linux
Pull string hash improvements from George Spelvin: "This series does several related things: - Makes the dcache hash (fs/namei.c) useful for general kernel use. (Thanks to Bruce for noticing the zero-length corner case) - Converts the string hashes in <linux/sunrpc/svcauth.h> to use the above. - Avoids 64-bit multiplies in hash_64() on 32-bit platforms. Two 32-bit multiplies will do well enough. - Rids the world of the bad hash multipliers in hash_32. This finishes the job started in commit 689de1d6ca95 ("Minimal fix-up of bad hashing behavior of hash_64()") The vast majority of Linux architectures have hardware support for 32x32-bit multiply and so derive no benefit from "simplified" multipliers. The few processors that do not (68000, h8/300 and some models of Microblaze) have arch-specific implementations added. Those patches are last in the series. - Overhauls the dcache hash mixing. The patch in commit 0fed3ac866ea ("namei: Improve hash mixing if CONFIG_DCACHE_WORD_ACCESS") was an off-the-cuff suggestion. Replaced with a much more careful design that's simultaneously faster and better. (My own invention, as there was noting suitable in the literature I could find. Comments welcome!) - Modify the hash_name() loop to skip the initial HASH_MIX(). This would let us salt the hash if we ever wanted to. - Sort out partial_name_hash(). The hash function is declared as using a long state, even though it's truncated to 32 bits at the end and the extra internal state contributes nothing to the result. And some callers do odd things: - fs/hfs/string.c only allocates 32 bits of state - fs/hfsplus/unicode.c uses it to hash 16-bit unicode symbols not bytes - Modify bytemask_from_count to handle inputs of 1..sizeof(long) rather than 0..sizeof(long)-1. This would simplify users other than full_name_hash" Special thanks to Bruce Fields for testing and finding bugs in v1. (I learned some humbling lessons about "obviously correct" code.) On the arch-specific front, the m68k assembly has been tested in a standalone test harness, I've been in contact with the Microblaze maintainers who mostly don't care, as the hardware multiplier is never omitted in real-world applications, and I haven't heard anything from the H8/300 world" * 'hash' of git://ftp.sciencehorizons.net/linux: h8300: Add <asm/hash.h> microblaze: Add <asm/hash.h> m68k: Add <asm/hash.h> <linux/hash.h>: Add support for architecture-specific functions fs/namei.c: Improve dcache hash function Eliminate bad hash multipliers from hash_32() and hash_64() Change hash_64() return value to 32 bits <linux/sunrpc/svcauth.h>: Define hash_str() in terms of hashlen_string() fs/namei.c: Add hashlen_string() function Pull out string hash to <linux/stringhash.h>
Diffstat (limited to 'lib')
-rw-r--r--lib/Kconfig.debug11
-rw-r--r--lib/Makefile1
-rw-r--r--lib/test_hash.c250
3 files changed, 262 insertions, 0 deletions
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index e707ab3e1991..77d7d034bac3 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1849,6 +1849,17 @@ config TEST_RHASHTABLE
If unsure, say N.
+config TEST_HASH
+ tristate "Perform selftest on hash functions"
+ default n
+ help
+ Enable this option to test the kernel's integer (<linux/hash,h>)
+ and string (<linux/stringhash.h>) hash functions on boot
+ (or module load).
+
+ This is intended to help people writing architecture-specific
+ optimized versions. If unsure, say N.
+
endmenu # runtime tests
config PROVIDE_OHCI1394_DMA_INIT
diff --git a/lib/Makefile b/lib/Makefile
index 42b69185f963..499fb354d627 100644
--- a/lib/Makefile
+++ b/lib/Makefile
@@ -48,6 +48,7 @@ obj-$(CONFIG_TEST_HEXDUMP) += test_hexdump.o
obj-y += kstrtox.o
obj-$(CONFIG_TEST_BPF) += test_bpf.o
obj-$(CONFIG_TEST_FIRMWARE) += test_firmware.o
+obj-$(CONFIG_TEST_HASH) += test_hash.o
obj-$(CONFIG_TEST_KASAN) += test_kasan.o
obj-$(CONFIG_TEST_KSTRTOX) += test-kstrtox.o
obj-$(CONFIG_TEST_LKM) += test_module.o
diff --git a/lib/test_hash.c b/lib/test_hash.c
new file mode 100644
index 000000000000..c9549c8b4909
--- /dev/null
+++ b/lib/test_hash.c
@@ -0,0 +1,250 @@
+/*
+ * Test cases for <linux/hash.h> and <linux/stringhash.h>
+ * This just verifies that various ways of computing a hash
+ * produce the same thing and, for cases where a k-bit hash
+ * value is requested, is of the requested size.
+ *
+ * We fill a buffer with a 255-byte null-terminated string,
+ * and use both full_name_hash() and hashlen_string() to hash the
+ * substrings from i to j, where 0 <= i < j < 256.
+ *
+ * The returned values are used to check that __hash_32() and
+ * __hash_32_generic() compute the same thing. Likewise hash_32()
+ * and hash_64().
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt "\n"
+
+#include <linux/compiler.h>
+#include <linux/types.h>
+#include <linux/module.h>
+#include <linux/hash.h>
+#include <linux/stringhash.h>
+#include <linux/printk.h>
+
+/* 32-bit XORSHIFT generator. Seed must not be zero. */
+static u32 __init __attribute_const__
+xorshift(u32 seed)
+{
+ seed ^= seed << 13;
+ seed ^= seed >> 17;
+ seed ^= seed << 5;
+ return seed;
+}
+
+/* Given a non-zero x, returns a non-zero byte. */
+static u8 __init __attribute_const__
+mod255(u32 x)
+{
+ x = (x & 0xffff) + (x >> 16); /* 1 <= x <= 0x1fffe */
+ x = (x & 0xff) + (x >> 8); /* 1 <= x <= 0x2fd */
+ x = (x & 0xff) + (x >> 8); /* 1 <= x <= 0x100 */
+ x = (x & 0xff) + (x >> 8); /* 1 <= x <= 0xff */
+ return x;
+}
+
+/* Fill the buffer with non-zero bytes. */
+static void __init
+fill_buf(char *buf, size_t len, u32 seed)
+{
+ size_t i;
+
+ for (i = 0; i < len; i++) {
+ seed = xorshift(seed);
+ buf[i] = mod255(seed);
+ }
+}
+
+/*
+ * Test the various integer hash functions. h64 (or its low-order bits)
+ * is the integer to hash. hash_or accumulates the OR of the hash values,
+ * which are later checked to see that they cover all the requested bits.
+ *
+ * Because these functions (as opposed to the string hashes) are all
+ * inline, the code being tested is actually in the module, and you can
+ * recompile and re-test the module without rebooting.
+ */
+static bool __init
+test_int_hash(unsigned long long h64, u32 hash_or[2][33])
+{
+ int k;
+ u32 h0 = (u32)h64, h1, h2;
+
+ /* Test __hash32 */
+ hash_or[0][0] |= h1 = __hash_32(h0);
+#ifdef HAVE_ARCH__HASH_32
+ hash_or[1][0] |= h2 = __hash_32_generic(h0);
+#if HAVE_ARCH__HASH_32 == 1
+ if (h1 != h2) {
+ pr_err("__hash_32(%#x) = %#x != __hash_32_generic() = %#x",
+ h0, h1, h2);
+ return false;
+ }
+#endif
+#endif
+
+ /* Test k = 1..32 bits */
+ for (k = 1; k <= 32; k++) {
+ u32 const m = ((u32)2 << (k-1)) - 1; /* Low k bits set */
+
+ /* Test hash_32 */
+ hash_or[0][k] |= h1 = hash_32(h0, k);
+ if (h1 > m) {
+ pr_err("hash_32(%#x, %d) = %#x > %#x", h0, k, h1, m);
+ return false;
+ }
+#ifdef HAVE_ARCH_HASH_32
+ h2 = hash_32_generic(h0, k);
+#if HAVE_ARCH_HASH_32 == 1
+ if (h1 != h2) {
+ pr_err("hash_32(%#x, %d) = %#x != hash_32_generic() "
+ " = %#x", h0, k, h1, h2);
+ return false;
+ }
+#else
+ if (h2 > m) {
+ pr_err("hash_32_generic(%#x, %d) = %#x > %#x",
+ h0, k, h1, m);
+ return false;
+ }
+#endif
+#endif
+ /* Test hash_64 */
+ hash_or[1][k] |= h1 = hash_64(h64, k);
+ if (h1 > m) {
+ pr_err("hash_64(%#llx, %d) = %#x > %#x", h64, k, h1, m);
+ return false;
+ }
+#ifdef HAVE_ARCH_HASH_64
+ h2 = hash_64_generic(h64, k);
+#if HAVE_ARCH_HASH_64 == 1
+ if (h1 != h2) {
+ pr_err("hash_64(%#llx, %d) = %#x != hash_64_generic() "
+ "= %#x", h64, k, h1, h2);
+ return false;
+ }
+#else
+ if (h2 > m) {
+ pr_err("hash_64_generic(%#llx, %d) = %#x > %#x",
+ h64, k, h1, m);
+ return false;
+ }
+#endif
+#endif
+ }
+
+ (void)h2; /* Suppress unused variable warning */
+ return true;
+}
+
+#define SIZE 256 /* Run time is cubic in SIZE */
+
+static int __init
+test_hash_init(void)
+{
+ char buf[SIZE+1];
+ u32 string_or = 0, hash_or[2][33] = { 0 };
+ unsigned tests = 0;
+ unsigned long long h64 = 0;
+ int i, j;
+
+ fill_buf(buf, SIZE, 1);
+
+ /* Test every possible non-empty substring in the buffer. */
+ for (j = SIZE; j > 0; --j) {
+ buf[j] = '\0';
+
+ for (i = 0; i <= j; i++) {
+ u64 hashlen = hashlen_string(buf+i);
+ u32 h0 = full_name_hash(buf+i, j-i);
+
+ /* Check that hashlen_string gets the length right */
+ if (hashlen_len(hashlen) != j-i) {
+ pr_err("hashlen_string(%d..%d) returned length"
+ " %u, expected %d",
+ i, j, hashlen_len(hashlen), j-i);
+ return -EINVAL;
+ }
+ /* Check that the hashes match */
+ if (hashlen_hash(hashlen) != h0) {
+ pr_err("hashlen_string(%d..%d) = %08x != "
+ "full_name_hash() = %08x",
+ i, j, hashlen_hash(hashlen), h0);
+ return -EINVAL;
+ }
+
+ string_or |= h0;
+ h64 = h64 << 32 | h0; /* For use with hash_64 */
+ if (!test_int_hash(h64, hash_or))
+ return -EINVAL;
+ tests++;
+ } /* i */
+ } /* j */
+
+ /* The OR of all the hash values should cover all the bits */
+ if (~string_or) {
+ pr_err("OR of all string hash results = %#x != %#x",
+ string_or, -1u);
+ return -EINVAL;
+ }
+ if (~hash_or[0][0]) {
+ pr_err("OR of all __hash_32 results = %#x != %#x",
+ hash_or[0][0], -1u);
+ return -EINVAL;
+ }
+#ifdef HAVE_ARCH__HASH_32
+#if HAVE_ARCH__HASH_32 != 1 /* Test is pointless if results match */
+ if (~hash_or[1][0]) {
+ pr_err("OR of all __hash_32_generic results = %#x != %#x",
+ hash_or[1][0], -1u);
+ return -EINVAL;
+ }
+#endif
+#endif
+
+ /* Likewise for all the i-bit hash values */
+ for (i = 1; i <= 32; i++) {
+ u32 const m = ((u32)2 << (i-1)) - 1; /* Low i bits set */
+
+ if (hash_or[0][i] != m) {
+ pr_err("OR of all hash_32(%d) results = %#x "
+ "(%#x expected)", i, hash_or[0][i], m);
+ return -EINVAL;
+ }
+ if (hash_or[1][i] != m) {
+ pr_err("OR of all hash_64(%d) results = %#x "
+ "(%#x expected)", i, hash_or[1][i], m);
+ return -EINVAL;
+ }
+ }
+
+ /* Issue notices about skipped tests. */
+#ifndef HAVE_ARCH__HASH_32
+ pr_info("__hash_32() has no arch implementation to test.");
+#elif HAVE_ARCH__HASH_32 != 1
+ pr_info("__hash_32() is arch-specific; not compared to generic.");
+#endif
+#ifndef HAVE_ARCH_HASH_32
+ pr_info("hash_32() has no arch implementation to test.");
+#elif HAVE_ARCH_HASH_32 != 1
+ pr_info("hash_32() is arch-specific; not compared to generic.");
+#endif
+#ifndef HAVE_ARCH_HASH_64
+ pr_info("hash_64() has no arch implementation to test.");
+#elif HAVE_ARCH_HASH_64 != 1
+ pr_info("hash_64() is arch-specific; not compared to generic.");
+#endif
+
+ pr_notice("%u tests passed.", tests);
+
+ return 0;
+}
+
+static void __exit test_hash_exit(void)
+{
+}
+
+module_init(test_hash_init); /* Does everything */
+module_exit(test_hash_exit); /* Does nothing */
+
+MODULE_LICENSE("GPL");