pf: Add a new zone for per-table entry counters.

Right now we optionally allocate 8 counters per table entry, so in
addition to memory consumed by counters, we require 8 pointers worth of
space in each entry even when counters are not allocated (the default).

Instead, define a UMA zone that returns contiguous per-CPU counter
arrays for use in table entries.  On amd64 this reduces sizeof(struct
pfr_kentry) from 216 to 160.  The smaller size also results in better
slab efficiency, so memory usage for large tables is reduced by about
28%.

Reviewed by:	kp
MFC after:	2 weeks
Differential Revision:	https://reviews.freebsd.org/D24843
This commit is contained in:
markj 2020-05-16 00:28:12 +00:00
parent f152004cfe
commit 68806c4044
2 changed files with 37 additions and 71 deletions

View File

@ -1000,6 +1000,8 @@ struct pfr_addr {
enum { PFR_DIR_IN, PFR_DIR_OUT, PFR_DIR_MAX };
enum { PFR_OP_BLOCK, PFR_OP_PASS, PFR_OP_ADDR_MAX, PFR_OP_TABLE_MAX };
enum { PFR_TYPE_PACKETS, PFR_TYPE_BYTES, PFR_TYPE_MAX };
#define PFR_NUM_COUNTERS (PFR_DIR_MAX * PFR_OP_ADDR_MAX * PFR_TYPE_MAX)
#define PFR_OP_XPASS PFR_OP_ADDR_MAX
struct pfr_astats {
@ -1045,10 +1047,12 @@ union sockaddr_union {
#endif /* _SOCKADDR_UNION_DEFINED */
struct pfr_kcounters {
counter_u64_t pfrkc_packets[PFR_DIR_MAX][PFR_OP_ADDR_MAX];
counter_u64_t pfrkc_bytes[PFR_DIR_MAX][PFR_OP_ADDR_MAX];
counter_u64_t pfrkc_counters;
long pfrkc_tzero;
};
#define pfr_kentry_counter(kc, dir, op, t) \
((kc)->pfrkc_counters + \
(dir) * PFR_OP_ADDR_MAX * PFR_TYPE_MAX + (op) * PFR_TYPE_MAX + (t))
SLIST_HEAD(pfr_kentryworkq, pfr_kentry);
struct pfr_kentry {

View File

@ -127,6 +127,8 @@ struct pfr_walktree {
static MALLOC_DEFINE(M_PFTABLE, "pf_table", "pf(4) tables structures");
VNET_DEFINE_STATIC(uma_zone_t, pfr_kentry_z);
#define V_pfr_kentry_z VNET(pfr_kentry_z)
VNET_DEFINE_STATIC(uma_zone_t, pfr_kentry_counter_z);
#define V_pfr_kentry_counter_z VNET(pfr_kentry_counter_z)
static struct pf_addr pfr_ffaddr = {
.addr32 = { 0xffffffff, 0xffffffff, 0xffffffff, 0xffffffff }
@ -144,12 +146,8 @@ static void pfr_mark_addrs(struct pfr_ktable *);
static struct pfr_kentry
*pfr_lookup_addr(struct pfr_ktable *,
struct pfr_addr *, int);
static bool pfr_create_kentry_counter(struct pfr_kentry *, int,
int);
static struct pfr_kentry *pfr_create_kentry(struct pfr_addr *, bool);
static void pfr_destroy_kentries(struct pfr_kentryworkq *);
static void pfr_destroy_kentry_counter(struct pfr_kcounters *,
int, int);
static void pfr_destroy_kentry(struct pfr_kentry *);
static void pfr_insert_kentries(struct pfr_ktable *,
struct pfr_kentryworkq *, long);
@ -205,6 +203,9 @@ void
pfr_initialize(void)
{
V_pfr_kentry_counter_z = uma_zcreate("pf table entry counters",
PFR_NUM_COUNTERS * sizeof(uint64_t), NULL, NULL, NULL, NULL,
UMA_ALIGN_PTR, UMA_ZONE_PCPU);
V_pfr_kentry_z = uma_zcreate("pf table entries",
sizeof(struct pfr_kentry), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR,
0);
@ -217,6 +218,7 @@ pfr_cleanup(void)
{
uma_zdestroy(V_pfr_kentry_z);
uma_zdestroy(V_pfr_kentry_counter_z);
}
int
@ -785,27 +787,11 @@ pfr_lookup_addr(struct pfr_ktable *kt, struct pfr_addr *ad, int exact)
return (ke);
}
static bool
pfr_create_kentry_counter(struct pfr_kentry *ke, int pfr_dir, int pfr_op)
{
counter_u64_t c;
c = counter_u64_alloc(M_NOWAIT);
if (c == NULL)
return (false);
ke->pfrke_counters.pfrkc_packets[pfr_dir][pfr_op] = c;
c = counter_u64_alloc(M_NOWAIT);
if (c == NULL)
return (false);
ke->pfrke_counters.pfrkc_bytes[pfr_dir][pfr_op] = c;
return (true);
}
static struct pfr_kentry *
pfr_create_kentry(struct pfr_addr *ad, bool counters)
{
struct pfr_kentry *ke;
int pfr_dir, pfr_op;
counter_u64_t c;
ke = uma_zalloc(V_pfr_kentry_z, M_NOWAIT | M_ZERO);
if (ke == NULL)
@ -819,15 +805,14 @@ pfr_create_kentry(struct pfr_addr *ad, bool counters)
ke->pfrke_net = ad->pfra_net;
ke->pfrke_not = ad->pfra_not;
ke->pfrke_counters.pfrkc_tzero = 0;
if (counters)
for (pfr_dir = 0; pfr_dir < PFR_DIR_MAX; pfr_dir++)
for (pfr_op = 0; pfr_op < PFR_OP_ADDR_MAX; pfr_op++) {
if (!pfr_create_kentry_counter(ke, pfr_dir,
pfr_op)) {
pfr_destroy_kentry(ke);
return (NULL);
}
}
if (counters) {
c = uma_zalloc_pcpu(V_pfr_kentry_counter_z, M_NOWAIT | M_ZERO);
if (c == NULL) {
pfr_destroy_kentry(ke);
return (NULL);
}
ke->pfrke_counters.pfrkc_counters = c;
}
return (ke);
}
@ -843,26 +828,12 @@ pfr_destroy_kentries(struct pfr_kentryworkq *workq)
}
static void
pfr_destroy_kentry_counter(struct pfr_kcounters *kc, int pfr_dir, int pfr_op)
pfr_destroy_kentry(struct pfr_kentry *ke)
{
counter_u64_t c;
if ((c = kc->pfrkc_packets[pfr_dir][pfr_op]) != NULL)
counter_u64_free(c);
if ((c = kc->pfrkc_bytes[pfr_dir][pfr_op]) != NULL)
counter_u64_free(c);
}
static void
pfr_destroy_kentry(struct pfr_kentry *ke)
{
int pfr_dir, pfr_op;
for (pfr_dir = 0; pfr_dir < PFR_DIR_MAX; pfr_dir ++)
for (pfr_op = 0; pfr_op < PFR_OP_ADDR_MAX; pfr_op ++)
pfr_destroy_kentry_counter(&ke->pfrke_counters,
pfr_dir, pfr_op);
if ((c = ke->pfrke_counters.pfrkc_counters) != NULL)
uma_zfree_pcpu(V_pfr_kentry_counter_z, c);
uma_zfree(V_pfr_kentry_z, ke);
}
@ -934,29 +905,20 @@ pfr_clean_node_mask(struct pfr_ktable *kt,
pfr_unroute_kentry(kt, p);
}
static void
pfr_clear_kentry_counters(struct pfr_kentry *p, int pfr_dir, int pfr_op)
{
counter_u64_zero(p->pfrke_counters.pfrkc_packets[pfr_dir][pfr_op]);
counter_u64_zero(p->pfrke_counters.pfrkc_bytes[pfr_dir][pfr_op]);
}
static void
pfr_clstats_kentries(struct pfr_ktable *kt, struct pfr_kentryworkq *workq,
long tzero, int negchange)
{
struct pfr_kentry *p;
int pfr_dir, pfr_op;
int i;
SLIST_FOREACH(p, workq, pfrke_workq) {
if (negchange)
p->pfrke_not = !p->pfrke_not;
if ((kt->pfrkt_flags & PFR_TFLAG_COUNTERS) != 0)
for (pfr_dir = 0; pfr_dir < PFR_DIR_MAX; pfr_dir++)
for (pfr_op = 0; pfr_op < PFR_OP_ADDR_MAX;
pfr_op++)
pfr_clear_kentry_counters(p, pfr_dir,
pfr_op);
for (i = 0; i < PFR_NUM_COUNTERS; i++)
counter_u64_zero(
p->pfrke_counters.pfrkc_counters + i);
p->pfrke_counters.pfrkc_tzero = tzero;
}
}
@ -1077,12 +1039,12 @@ pfr_copyout_astats(struct pfr_astats *as, const struct pfr_kentry *ke,
return;
}
for (dir = 0; dir < PFR_DIR_MAX; dir ++) {
for (dir = 0; dir < PFR_DIR_MAX; dir++) {
for (op = 0; op < PFR_OP_ADDR_MAX; op ++) {
as->pfras_packets[dir][op] =
counter_u64_fetch(kc->pfrkc_packets[dir][op]);
as->pfras_bytes[dir][op] =
counter_u64_fetch(kc->pfrkc_bytes[dir][op]);
as->pfras_packets[dir][op] = counter_u64_fetch(
pfr_kentry_counter(kc, dir, op, PFR_TYPE_PACKETS));
as->pfras_bytes[dir][op] = counter_u64_fetch(
pfr_kentry_counter(kc, dir, op, PFR_TYPE_BYTES));
}
}
}
@ -2147,10 +2109,10 @@ pfr_update_stats(struct pfr_ktable *kt, struct pf_addr *a, sa_family_t af,
counter_u64_add(kt->pfrkt_bytes[dir_out][op_pass], len);
if (ke != NULL && op_pass != PFR_OP_XPASS &&
(kt->pfrkt_flags & PFR_TFLAG_COUNTERS)) {
counter_u64_add(ke->pfrke_counters.
pfrkc_packets[dir_out][op_pass], 1);
counter_u64_add(ke->pfrke_counters.
pfrkc_bytes[dir_out][op_pass], len);
counter_u64_add(pfr_kentry_counter(&ke->pfrke_counters,
dir_out, op_pass, PFR_TYPE_PACKETS), 1);
counter_u64_add(pfr_kentry_counter(&ke->pfrke_counters,
dir_out, op_pass, PFR_TYPE_BYTES), len);
}
}