From: Pablo Neira Ayuso <pablo@netfilter.org>
To: netfilter-devel@vger.kernel.org
Cc: davem@davemloft.net, netdev@vger.kernel.org, kuba@kernel.org,
pabeni@redhat.com, edumazet@google.com, fw@strlen.de
Subject: [PATCH net-next 02/16] netfilter: nft_counter: Use u64_stats_t for statistic.
Date: Fri, 6 Sep 2024 01:29:06 +0200 [thread overview]
Message-ID: <20240905232920.5481-3-pablo@netfilter.org> (raw)
In-Reply-To: <20240905232920.5481-1-pablo@netfilter.org>
From: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
The nft_counter uses two s64 counters for statistics. Those two are
protected by a seqcount to ensure that the 64bit variable is always
properly seen during updates even on 32bit architectures where the store
is performed by two writes. A side effect is that the two counter (bytes
and packet) are written and read together in the same window.
This can be replaced with u64_stats_t. write_seqcount_begin()/ end() is
replaced with u64_stats_update_begin()/ end() and behaves the same way
as with seqcount_t on 32bit architectures. Additionally there is a
preempt_disable on PREEMPT_RT to ensure that a reader does not preempt a
writer.
On 64bit architectures the macros are removed and the reads happen
without any retries. This also means that the reader can observe one
counter (bytes) from before the update and the other counter (packets)
but that is okay since there is no requirement to have both counter from
the same update window.
Convert the statistic to u64_stats_t. There is one optimisation:
nft_counter_do_init() and nft_counter_clone() allocate a new per-CPU
counter and assign a value to it. During this assignment preemption is
disabled which is not needed because the counter is not yet exposed to
the system so there can not be another writer or reader. Therefore
disabling preemption is omitted and raw_cpu_ptr() is used to obtain a
pointer to a counter for the assignment.
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
net/netfilter/nft_counter.c | 90 +++++++++++++++++++------------------
1 file changed, 46 insertions(+), 44 deletions(-)
diff --git a/net/netfilter/nft_counter.c b/net/netfilter/nft_counter.c
index eab0dc66bee6..cc7325329496 100644
--- a/net/netfilter/nft_counter.c
+++ b/net/netfilter/nft_counter.c
@@ -8,7 +8,7 @@
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/module.h>
-#include <linux/seqlock.h>
+#include <linux/u64_stats_sync.h>
#include <linux/netlink.h>
#include <linux/netfilter.h>
#include <linux/netfilter/nf_tables.h>
@@ -17,6 +17,11 @@
#include <net/netfilter/nf_tables_offload.h>
struct nft_counter {
+ u64_stats_t bytes;
+ u64_stats_t packets;
+};
+
+struct nft_counter_tot {
s64 bytes;
s64 packets;
};
@@ -25,25 +30,24 @@ struct nft_counter_percpu_priv {
struct nft_counter __percpu *counter;
};
-static DEFINE_PER_CPU(seqcount_t, nft_counter_seq);
+static DEFINE_PER_CPU(struct u64_stats_sync, nft_counter_sync);
static inline void nft_counter_do_eval(struct nft_counter_percpu_priv *priv,
struct nft_regs *regs,
const struct nft_pktinfo *pkt)
{
+ struct u64_stats_sync *nft_sync;
struct nft_counter *this_cpu;
- seqcount_t *myseq;
local_bh_disable();
this_cpu = this_cpu_ptr(priv->counter);
- myseq = this_cpu_ptr(&nft_counter_seq);
-
- write_seqcount_begin(myseq);
+ nft_sync = this_cpu_ptr(&nft_counter_sync);
- this_cpu->bytes += pkt->skb->len;
- this_cpu->packets++;
+ u64_stats_update_begin(nft_sync);
+ u64_stats_add(&this_cpu->bytes, pkt->skb->len);
+ u64_stats_inc(&this_cpu->packets);
+ u64_stats_update_end(nft_sync);
- write_seqcount_end(myseq);
local_bh_enable();
}
@@ -66,17 +70,16 @@ static int nft_counter_do_init(const struct nlattr * const tb[],
if (cpu_stats == NULL)
return -ENOMEM;
- preempt_disable();
- this_cpu = this_cpu_ptr(cpu_stats);
+ this_cpu = raw_cpu_ptr(cpu_stats);
if (tb[NFTA_COUNTER_PACKETS]) {
- this_cpu->packets =
- be64_to_cpu(nla_get_be64(tb[NFTA_COUNTER_PACKETS]));
+ u64_stats_set(&this_cpu->packets,
+ be64_to_cpu(nla_get_be64(tb[NFTA_COUNTER_PACKETS])));
}
if (tb[NFTA_COUNTER_BYTES]) {
- this_cpu->bytes =
- be64_to_cpu(nla_get_be64(tb[NFTA_COUNTER_BYTES]));
+ u64_stats_set(&this_cpu->bytes,
+ be64_to_cpu(nla_get_be64(tb[NFTA_COUNTER_BYTES])));
}
- preempt_enable();
+
priv->counter = cpu_stats;
return 0;
}
@@ -104,40 +107,41 @@ static void nft_counter_obj_destroy(const struct nft_ctx *ctx,
}
static void nft_counter_reset(struct nft_counter_percpu_priv *priv,
- struct nft_counter *total)
+ struct nft_counter_tot *total)
{
+ struct u64_stats_sync *nft_sync;
struct nft_counter *this_cpu;
- seqcount_t *myseq;
local_bh_disable();
this_cpu = this_cpu_ptr(priv->counter);
- myseq = this_cpu_ptr(&nft_counter_seq);
+ nft_sync = this_cpu_ptr(&nft_counter_sync);
+
+ u64_stats_update_begin(nft_sync);
+ u64_stats_add(&this_cpu->packets, -total->packets);
+ u64_stats_add(&this_cpu->bytes, -total->bytes);
+ u64_stats_update_end(nft_sync);
- write_seqcount_begin(myseq);
- this_cpu->packets -= total->packets;
- this_cpu->bytes -= total->bytes;
- write_seqcount_end(myseq);
local_bh_enable();
}
static void nft_counter_fetch(struct nft_counter_percpu_priv *priv,
- struct nft_counter *total)
+ struct nft_counter_tot *total)
{
struct nft_counter *this_cpu;
- const seqcount_t *myseq;
u64 bytes, packets;
unsigned int seq;
int cpu;
memset(total, 0, sizeof(*total));
for_each_possible_cpu(cpu) {
- myseq = per_cpu_ptr(&nft_counter_seq, cpu);
+ struct u64_stats_sync *nft_sync = per_cpu_ptr(&nft_counter_sync, cpu);
+
this_cpu = per_cpu_ptr(priv->counter, cpu);
do {
- seq = read_seqcount_begin(myseq);
- bytes = this_cpu->bytes;
- packets = this_cpu->packets;
- } while (read_seqcount_retry(myseq, seq));
+ seq = u64_stats_fetch_begin(nft_sync);
+ bytes = u64_stats_read(&this_cpu->bytes);
+ packets = u64_stats_read(&this_cpu->packets);
+ } while (u64_stats_fetch_retry(nft_sync, seq));
total->bytes += bytes;
total->packets += packets;
@@ -148,7 +152,7 @@ static int nft_counter_do_dump(struct sk_buff *skb,
struct nft_counter_percpu_priv *priv,
bool reset)
{
- struct nft_counter total;
+ struct nft_counter_tot total;
nft_counter_fetch(priv, &total);
@@ -237,7 +241,7 @@ static int nft_counter_clone(struct nft_expr *dst, const struct nft_expr *src, g
struct nft_counter_percpu_priv *priv_clone = nft_expr_priv(dst);
struct nft_counter __percpu *cpu_stats;
struct nft_counter *this_cpu;
- struct nft_counter total;
+ struct nft_counter_tot total;
nft_counter_fetch(priv, &total);
@@ -245,11 +249,9 @@ static int nft_counter_clone(struct nft_expr *dst, const struct nft_expr *src, g
if (cpu_stats == NULL)
return -ENOMEM;
- preempt_disable();
- this_cpu = this_cpu_ptr(cpu_stats);
- this_cpu->packets = total.packets;
- this_cpu->bytes = total.bytes;
- preempt_enable();
+ this_cpu = raw_cpu_ptr(cpu_stats);
+ u64_stats_set(&this_cpu->packets, total.packets);
+ u64_stats_set(&this_cpu->bytes, total.bytes);
priv_clone->counter = cpu_stats;
return 0;
@@ -267,17 +269,17 @@ static void nft_counter_offload_stats(struct nft_expr *expr,
const struct flow_stats *stats)
{
struct nft_counter_percpu_priv *priv = nft_expr_priv(expr);
+ struct u64_stats_sync *nft_sync;
struct nft_counter *this_cpu;
- seqcount_t *myseq;
local_bh_disable();
this_cpu = this_cpu_ptr(priv->counter);
- myseq = this_cpu_ptr(&nft_counter_seq);
+ nft_sync = this_cpu_ptr(&nft_counter_sync);
- write_seqcount_begin(myseq);
- this_cpu->packets += stats->pkts;
- this_cpu->bytes += stats->bytes;
- write_seqcount_end(myseq);
+ u64_stats_update_begin(nft_sync);
+ u64_stats_add(&this_cpu->packets, stats->pkts);
+ u64_stats_add(&this_cpu->bytes, stats->bytes);
+ u64_stats_update_end(nft_sync);
local_bh_enable();
}
@@ -286,7 +288,7 @@ void nft_counter_init_seqcount(void)
int cpu;
for_each_possible_cpu(cpu)
- seqcount_init(per_cpu_ptr(&nft_counter_seq, cpu));
+ u64_stats_init(per_cpu_ptr(&nft_counter_sync, cpu));
}
struct nft_expr_type nft_counter_type;
--
2.30.2
next prev parent reply other threads:[~2024-09-05 23:29 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-09-05 23:29 [PATCH net-next 00/16] Netfilter updates for net-next Pablo Neira Ayuso
2024-09-05 23:29 ` [PATCH net-next 01/16] netfilter: ctnetlink: support CTA_FILTER for flush Pablo Neira Ayuso
2024-09-07 2:10 ` patchwork-bot+netdevbpf
2024-09-05 23:29 ` Pablo Neira Ayuso [this message]
2024-09-05 23:29 ` [PATCH net-next 03/16] netfilter: Use kmemdup_array instead of kmemdup for multiple allocation Pablo Neira Ayuso
2024-09-05 23:29 ` [PATCH net-next 04/16] netfilter: conntrack: Convert to use ERR_CAST() Pablo Neira Ayuso
2024-09-05 23:29 ` [PATCH net-next 05/16] netfilter: nf_tables: drop unused 3rd argument from validate callback ops Pablo Neira Ayuso
2024-09-05 23:29 ` [PATCH net-next 06/16] netfilter: nf_tables: Correct spelling in nf_tables.h Pablo Neira Ayuso
2024-09-05 23:29 ` [PATCH net-next 07/16] netfilter: nf_tables: Add missing Kernel doc Pablo Neira Ayuso
2024-09-05 23:29 ` [PATCH net-next 08/16] netfilter: nf_tables: elements with timeout below CONFIG_HZ never expire Pablo Neira Ayuso
2024-09-05 23:29 ` [PATCH net-next 09/16] netfilter: nf_tables: reject element expiration with no timeout Pablo Neira Ayuso
2024-09-05 23:29 ` [PATCH net-next 10/16] netfilter: nf_tables: reject expiration higher than timeout Pablo Neira Ayuso
2024-09-05 23:29 ` [PATCH net-next 11/16] netfilter: nf_tables: remove annotation to access set timeout while holding lock Pablo Neira Ayuso
2024-09-05 23:29 ` [PATCH net-next 12/16] netfilter: nft_dynset: annotate data-races around set timeout Pablo Neira Ayuso
2024-09-05 23:29 ` [PATCH net-next 13/16] netfilter: nf_tables: annotate data-races around element expiration Pablo Neira Ayuso
2024-09-05 23:29 ` [PATCH net-next 14/16] netfilter: nf_tables: consolidate timeout extension for elements Pablo Neira Ayuso
2024-09-05 23:29 ` [PATCH net-next 15/16] netfilter: nf_tables: zero timeout means element never times out Pablo Neira Ayuso
2024-09-05 23:29 ` [PATCH net-next 16/16] netfilter: nf_tables: set element timeout update support Pablo Neira Ayuso
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20240905232920.5481-3-pablo@netfilter.org \
--to=pablo@netfilter.org \
--cc=davem@davemloft.net \
--cc=edumazet@google.com \
--cc=fw@strlen.de \
--cc=kuba@kernel.org \
--cc=netdev@vger.kernel.org \
--cc=netfilter-devel@vger.kernel.org \
--cc=pabeni@redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).