kernel_optimize_test/security/selinux/ss/hashtab.c
Ondrej Mosnacek 54b27f9287 selinux: complete the inlining of hashtab functions
Move (most of) the definitions of hashtab_search() and hashtab_insert()
to the header file. In combination with the previous patch, this avoids
calling the callbacks indirectly by function pointers and allows for
better optimization, leading to a drastic performance improvement of
these operations.

With this patch, I measured a speed up in the following areas (measured
on x86_64 F32 VM with 4 CPUs):
  1. Policy load (`load_policy`) - takes ~150 ms instead of ~230 ms.
  2. `chcon -R unconfined_u:object_r:user_tmp_t:s0:c381,c519 /tmp/linux-src`
     where /tmp/linux-src is an extracted linux-5.7 source tarball -
     takes ~522 ms instead of ~576 ms. This is because of many
     symtab_search() calls in string_to_context_struct() when there are
     many categories specified in the context.
  3. `stress-ng --msg 1 --msg-ops 10000000` - takes 12.41 s instead of
     13.95 s (consumes 18.6 s of kernel CPU time instead of 21.6 s).
     This is thanks to security_transition_sid() being ~43% faster after
     this patch.

Signed-off-by: Ondrej Mosnacek <omosnace@redhat.com>
Acked-by: Stephen Smalley <stephen.smalley.work@gmail.com>
Signed-off-by: Paul Moore <paul@paul-moore.com>
2020-07-09 19:08:16 -04:00

131 lines
2.8 KiB
C

// SPDX-License-Identifier: GPL-2.0
/*
* Implementation of the hash table type.
*
* Author : Stephen Smalley, <sds@tycho.nsa.gov>
*/
#include <linux/kernel.h>
#include <linux/slab.h>
#include <linux/errno.h>
#include "hashtab.h"
static struct kmem_cache *hashtab_node_cachep;
/*
* Here we simply round the number of elements up to the nearest power of two.
* I tried also other options like rouding down or rounding to the closest
* power of two (up or down based on which is closer), but I was unable to
* find any significant difference in lookup/insert performance that would
* justify switching to a different (less intuitive) formula. It could be that
* a different formula is actually more optimal, but any future changes here
* should be supported with performance/memory usage data.
*
* The total memory used by the htable arrays (only) with Fedora policy loaded
* is approximately 163 KB at the time of writing.
*/
static u32 hashtab_compute_size(u32 nel)
{
return nel == 0 ? 0 : roundup_pow_of_two(nel);
}
int hashtab_init(struct hashtab *h, u32 nel_hint)
{
h->size = hashtab_compute_size(nel_hint);
h->nel = 0;
if (!h->size)
return 0;
h->htable = kcalloc(h->size, sizeof(*h->htable), GFP_KERNEL);
return h->htable ? 0 : -ENOMEM;
}
int __hashtab_insert(struct hashtab *h, struct hashtab_node **dst,
void *key, void *datum)
{
struct hashtab_node *newnode;
newnode = kmem_cache_zalloc(hashtab_node_cachep, GFP_KERNEL);
if (!newnode)
return -ENOMEM;
newnode->key = key;
newnode->datum = datum;
newnode->next = *dst;
*dst = newnode;
h->nel++;
return 0;
}
void hashtab_destroy(struct hashtab *h)
{
u32 i;
struct hashtab_node *cur, *temp;
for (i = 0; i < h->size; i++) {
cur = h->htable[i];
while (cur) {
temp = cur;
cur = cur->next;
kmem_cache_free(hashtab_node_cachep, temp);
}
h->htable[i] = NULL;
}
kfree(h->htable);
h->htable = NULL;
}
int hashtab_map(struct hashtab *h,
int (*apply)(void *k, void *d, void *args),
void *args)
{
u32 i;
int ret;
struct hashtab_node *cur;
for (i = 0; i < h->size; i++) {
cur = h->htable[i];
while (cur) {
ret = apply(cur->key, cur->datum, args);
if (ret)
return ret;
cur = cur->next;
}
}
return 0;
}
void hashtab_stat(struct hashtab *h, struct hashtab_info *info)
{
u32 i, chain_len, slots_used, max_chain_len;
struct hashtab_node *cur;
slots_used = 0;
max_chain_len = 0;
for (i = 0; i < h->size; i++) {
cur = h->htable[i];
if (cur) {
slots_used++;
chain_len = 0;
while (cur) {
chain_len++;
cur = cur->next;
}
if (chain_len > max_chain_len)
max_chain_len = chain_len;
}
}
info->slots_used = slots_used;
info->max_chain_len = max_chain_len;
}
void __init hashtab_cache_init(void)
{
hashtab_node_cachep = kmem_cache_create("hashtab_node",
sizeof(struct hashtab_node),
0, SLAB_PANIC, NULL);
}