Merge af3228c69b into 1713aa7b4d

cmd/zstream/zstream_redup.c

@@ -132,7 +132,7 @@ static void
 rdt_insert(redup_table_t *rdt,
     uint64_t guid, uint64_t object, uint64_t offset, uint64_t stream_offset)
 {
-	uint64_t ch = cityhash4(guid, object, offset, 0);
+	uint64_t ch = cityhash3(guid, object, offset);
 	uint64_t hashcode = BF64_GET(ch, 0, rdt->numhashbits);
 	redup_entry_t **rdepp;
 
@@ -152,7 +152,7 @@ rdt_lookup(redup_table_t *rdt,
     uint64_t guid, uint64_t object, uint64_t offset,
     uint64_t *stream_offsetp)
 {
-	uint64_t ch = cityhash4(guid, object, offset, 0);
+	uint64_t ch = cityhash3(guid, object, offset);
 	uint64_t hashcode = BF64_GET(ch, 0, rdt->numhashbits);
 
 	for (redup_entry_t *rde = rdt->redup_hash_array[hashcode];

include/cityhash.h

@@ -32,6 +32,13 @@
 extern "C" {
 #endif
 
+/*
+ * Define 1/2/3-argument specialized versions of cityhash4, which can reduce
+ * instruction count (especially multiplication) on some 32-bit arches.
+ */
+_SYS_CITYHASH_H uint64_t cityhash1(uint64_t);
+_SYS_CITYHASH_H uint64_t cityhash2(uint64_t, uint64_t);
+_SYS_CITYHASH_H uint64_t cityhash3(uint64_t, uint64_t, uint64_t);
 _SYS_CITYHASH_H uint64_t cityhash4(uint64_t, uint64_t, uint64_t, uint64_t);
 
 #ifdef	__cplusplus

lib/libzfs/libzfs.abi

@@ -153,6 +153,9 @@
     <elf-symbol name='avl_update_lt' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
     <elf-symbol name='avl_walk' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
     <elf-symbol name='bookmark_namecheck' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
+    <elf-symbol name='cityhash1' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
+    <elf-symbol name='cityhash2' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
+    <elf-symbol name='cityhash3' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
     <elf-symbol name='cityhash4' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
     <elf-symbol name='color_end' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
     <elf-symbol name='color_start' type='func-type' binding='global-binding' visibility='default-visibility' is-defined='yes'/>
@@ -9179,6 +9182,21 @@
     </function-decl>
   </abi-instr>
   <abi-instr address-size='64' path='module/zcommon/cityhash.c' language='LANG_C99'>
+    <function-decl name='cityhash1' mangled-name='cityhash1' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='cityhash1'>
+      <parameter type-id='9c313c2d' name='w'/>
+      <return type-id='9c313c2d'/>
+    </function-decl>
+    <function-decl name='cityhash2' mangled-name='cityhash2' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='cityhash2'>
+      <parameter type-id='9c313c2d' name='w1'/>
+      <parameter type-id='9c313c2d' name='w2'/>
+      <return type-id='9c313c2d'/>
+    </function-decl>
+    <function-decl name='cityhash3' mangled-name='cityhash3' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='cityhash3'>
+      <parameter type-id='9c313c2d' name='w1'/>
+      <parameter type-id='9c313c2d' name='w2'/>
+      <parameter type-id='9c313c2d' name='w3'/>
+      <return type-id='9c313c2d'/>
+    </function-decl>
     <function-decl name='cityhash4' mangled-name='cityhash4' visibility='default' binding='global' size-in-bits='64' elf-symbol-id='cityhash4'>
       <parameter type-id='9c313c2d' name='w1'/>
       <parameter type-id='9c313c2d' name='w2'/>

module/os/linux/zfs/zvol_os.c

@@ -566,8 +566,8 @@ zvol_request_impl(zvol_state_t *zv, struct bio *bio, struct request *rq,
 		    rq->q->queue_hw_ctx[rq->q->mq_map[rq->cpu]]->queue_num;
 #endif
 #endif
-	taskq_hash = cityhash4((uintptr_t)zv, offset >> ZVOL_TASKQ_OFFSET_SHIFT,
+	taskq_hash = cityhash3((uintptr_t)zv, offset >> ZVOL_TASKQ_OFFSET_SHIFT,
-	    blk_mq_hw_queue, 0);
+	    blk_mq_hw_queue);
 	tq_idx = taskq_hash % ztqs->tqs_cnt;
 
 	if (rw == WRITE) {

module/zcommon/cityhash.c

@@ -49,8 +49,8 @@ cityhash_helper(uint64_t u, uint64_t v, uint64_t mul)
 	return (b);
 }
 
-uint64_t
+static inline uint64_t
-cityhash4(uint64_t w1, uint64_t w2, uint64_t w3, uint64_t w4)
+cityhash_impl(uint64_t w1, uint64_t w2, uint64_t w3, uint64_t w4)
 {
 	uint64_t mul = HASH_K2 + 64;
 	uint64_t a = w1 * HASH_K1;
@@ -59,9 +59,38 @@ cityhash4(uint64_t w1, uint64_t w2, uint64_t w3, uint64_t w4)
 	uint64_t d = w3 * HASH_K2;
 	return (cityhash_helper(rotate(a + b, 43) + rotate(c, 30) + d,
 	    a + rotate(b + HASH_K2, 18) + c, mul));
+}
 
+/*
+ * Passing w as the 2nd argument could save one 64-bit multiplication.
+ */
+uint64_t
+cityhash1(uint64_t w)
+{
+	return (cityhash_impl(0, w, 0, 0));
+}
+
+uint64_t
+cityhash2(uint64_t w1, uint64_t w2)
+{
+	return (cityhash_impl(w1, w2, 0, 0));
+}
+
+uint64_t
+cityhash3(uint64_t w1, uint64_t w2, uint64_t w3)
+{
+	return (cityhash_impl(w1, w2, w3, 0));
+}
+
+uint64_t
+cityhash4(uint64_t w1, uint64_t w2, uint64_t w3, uint64_t w4)
+{
+	return (cityhash_impl(w1, w2, w3, w4));
 }
 
 #if defined(_KERNEL)
+EXPORT_SYMBOL(cityhash1);
+EXPORT_SYMBOL(cityhash2);
+EXPORT_SYMBOL(cityhash3);
 EXPORT_SYMBOL(cityhash4);
 #endif

module/zfs/dmu_objset.c

@@ -66,6 +66,7 @@
 #include "zfs_namecheck.h"
 #include <sys/vdev_impl.h>
 #include <sys/arc.h>
+#include <cityhash.h>
 
 /*
  * Needed to close a window in dnode_move() that allows the objset to be freed
@@ -390,27 +391,13 @@ dmu_objset_byteswap(void *buf, size_t size)
 }
 
 /*
- * The hash is a CRC-based hash of the objset_t pointer and the object number.
+ * Runs cityhash on the objset_t pointer and the object number.
  */
 static uint64_t
 dnode_hash(const objset_t *os, uint64_t obj)
 {
 	uintptr_t osv = (uintptr_t)os;
-	uint64_t crc = -1ULL;
+	return (cityhash2((uint64_t)osv, obj));
-
-	ASSERT(zfs_crc64_table[128] == ZFS_CRC64_POLY);
-	/*
-	 * The lower 11 bits of the pointer don't have much entropy, because
-	 * the objset_t is more than 1KB long and so likely aligned to 2KB.
-	 */
-	crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (osv >> 11)) & 0xFF];
-	crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 0)) & 0xFF];
-	crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 8)) & 0xFF];
-	crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 16)) & 0xFF];
-
-	crc ^= (osv>>14) ^ (obj>>24);
-
-	return (crc);
 }
 
 static unsigned int

module/zfs/zio.c

@@ -4161,8 +4161,8 @@ zio_alloc_zil(spa_t *spa, objset_t *os, uint64_t txg, blkptr_t *new_bp,
 	 * some parallelism.
 	 */
 	int flags = METASLAB_ZIL;
-	int allocator = (uint_t)cityhash4(0, 0, 0,
+	int allocator = (uint_t)cityhash1(os->os_dsl_dataset->ds_object)
-	    os->os_dsl_dataset->ds_object) % spa->spa_alloc_count;
+	    % spa->spa_alloc_count;
 	error = metaslab_alloc(spa, spa_log_class(spa), size, new_bp, 1,
 	    txg, NULL, flags, &io_alloc_list, NULL, allocator);
 	*slog = (error == 0);