shared: switch our hash table implementation over to SipHash

SipHash appears to be the new gold standard for hashing smaller strings
for hashtables these days, so let's make use of it.

4 files changed, 25 insertions, 27 deletions

diff --git a/src/journal/catalog.c b/src/journal/catalog.c
index e3a3354ab..2823232cb 100644
--- a/src/journal/catalog.c
+++ b/src/journal/catalog.c
@@ -39,6 +39,7 @@
 #include "conf-files.h"
 #include "mkdir.h"
 #include "catalog.h"
+#include "siphash24.h"
 
 const char * const catalog_file_dirs[] = {
         "/usr/local/lib/systemd/catalog/",
@@ -63,28 +64,21 @@ typedef struct CatalogItem {
         le64_t offset;
 } CatalogItem;
 
-unsigned catalog_hash_func(const void *p) {
+unsigned long catalog_hash_func(const void *p, const uint8_t hash_key[HASH_KEY_SIZE]) {
         const CatalogItem *i = p;
+        uint64_t u;
+        size_t l, sz;
+        void *v;
 
-        assert_cc(sizeof(unsigned) == sizeof(uint8_t)*4);
-
-        return (((unsigned) i->id.bytes[0] << 24) |
-                ((unsigned) i->id.bytes[1] << 16) |
-                ((unsigned) i->id.bytes[2] << 8) |
-                ((unsigned) i->id.bytes[3])) ^
-               (((unsigned) i->id.bytes[4] << 24) |
-                ((unsigned) i->id.bytes[5] << 16) |
-                ((unsigned) i->id.bytes[6] << 8) |
-                ((unsigned) i->id.bytes[7])) ^
-               (((unsigned) i->id.bytes[8] << 24) |
-                ((unsigned) i->id.bytes[9] << 16) |
-                ((unsigned) i->id.bytes[10] << 8) |
-                ((unsigned) i->id.bytes[11])) ^
-               (((unsigned) i->id.bytes[12] << 24) |
-                ((unsigned) i->id.bytes[13] << 16) |
-                ((unsigned) i->id.bytes[14] << 8) |
-                ((unsigned) i->id.bytes[15])) ^
-                string_hash_func(i->language);
+        l = strlen(i->language);
+        sz = sizeof(i->id) + l;
+        v = alloca(sz);
+
+        memcpy(mempcpy(v, &i->id, sizeof(i->id)), i->language, l);
+
+        siphash24((uint8_t*) &u, v, sz, hash_key);
+
+        return (unsigned long) u;
 }
 
 int catalog_compare_func(const void *a, const void *b) {
diff --git a/src/journal/catalog.h b/src/journal/catalog.h
index 5e15b99ac..fdde67aee 100644
--- a/src/journal/catalog.h
+++ b/src/journal/catalog.h
@@ -28,7 +28,7 @@
 #include "strbuf.h"
 
 int catalog_import_file(Hashmap *h, struct strbuf *sb, const char *path);
-unsigned catalog_hash_func(const void *p);
+unsigned long catalog_hash_func(const void *p, const uint8_t hash_key[HASH_KEY_SIZE]);
 int catalog_compare_func(const void *a, const void *b) _pure_;
 int catalog_update(const char* database, const char* root, const char* const* dirs);
 int catalog_get(const char* database, sd_id128_t id, char **data);
diff --git a/src/journal/journal-file.c b/src/journal/journal-file.c
index b7e5cf0ab..121b40a58 100644
--- a/src/journal/journal-file.c
+++ b/src/journal/journal-file.c
@@ -2696,10 +2696,10 @@ int journal_file_open_reliably(
         /* The file is corrupted. Rotate it away and try it again (but only once) */
 
         l = strlen(fname);
-        if (asprintf(&p, "%.*s@%016llx-%016llx.journal~",
+        if (asprintf(&p, "%.*s@%016llx-%016" PRIx64 ".journal~",
                      (int) l - 8, fname,
                      (unsigned long long) now(CLOCK_REALTIME),
-                     random_ull()) < 0)
+                     random_u64()) < 0)
                 return -ENOMEM;
 
         r = rename(fname, p);
diff --git a/src/journal/journald-rate-limit.c b/src/journal/journald-rate-limit.c
index 4b7622152..6d779d296 100644
--- a/src/journal/journald-rate-limit.c
+++ b/src/journal/journald-rate-limit.c
@@ -56,7 +56,7 @@ struct JournalRateLimitGroup {
 
         char *id;
         JournalRateLimitPool pools[POOLS_MAX];
-        unsigned hash;
+        unsigned long hash;
 
         LIST_FIELDS(JournalRateLimitGroup, bucket);
         LIST_FIELDS(JournalRateLimitGroup, lru);
@@ -70,6 +70,8 @@ struct JournalRateLimit {
         JournalRateLimitGroup *lru, *lru_tail;
 
         unsigned n_groups;
+
+        uint8_t hash_key[16];
 };
 
 JournalRateLimit *journal_rate_limit_new(usec_t interval, unsigned burst) {
@@ -84,6 +86,8 @@ JournalRateLimit *journal_rate_limit_new(usec_t interval, unsigned burst) {
         r->interval = interval;
         r->burst = burst;
 
+        random_bytes(r->hash_key, sizeof(r->hash_key));
+
         return r;
 }
 
@@ -152,7 +156,7 @@ static JournalRateLimitGroup* journal_rate_limit_group_new(JournalRateLimit *r,
         if (!g->id)
                 goto fail;
 
-        g->hash = string_hash_func(g->id);
+        g->hash = string_hash_func(g->id, r->hash_key);
 
         journal_rate_limit_vacuum(r, ts);
 
@@ -199,7 +203,7 @@ static unsigned burst_modulate(unsigned burst, uint64_t available) {
 }
 
 int journal_rate_limit_test(JournalRateLimit *r, const char *id, int priority, uint64_t available) {
-        unsigned h;
+        unsigned long h;
         JournalRateLimitGroup *g;
         JournalRateLimitPool *p;
         unsigned burst;
@@ -217,7 +221,7 @@ int journal_rate_limit_test(JournalRateLimit *r, const char *id, int priority, u
 
         ts = now(CLOCK_MONOTONIC);
 
-        h = string_hash_func(id);
+        h = string_hash_func(id, r->hash_key);
         g = r->buckets[h % BUCKETS_MAX];
 
         LIST_FOREACH(bucket, g, g)