net/packet/Kconfig | 10 ++++++++++ net/packet/af_packet.c | 39 +++++++++++++++++++++++++++++++++++++++ net/packet/internal.h | 8 ++++++++ 3 files changed, 57 insertions(+)
In a system with high real-time requirements, the timeout mechanism of
ordinary timers with jiffies granularity is insufficient to meet the
demands for real-time performance. Meanwhile, the optimization of CPU
usage with af_packet is quite significant. Add hrtimer mode to help
compensate for the shortcomings in real-time performance.
Signed-off-by: Xin Zhao <jackzxcui1989@163.com>
---
net/packet/Kconfig | 10 ++++++++++
net/packet/af_packet.c | 39 +++++++++++++++++++++++++++++++++++++++
net/packet/internal.h | 8 ++++++++
3 files changed, 57 insertions(+)
diff --git a/net/packet/Kconfig b/net/packet/Kconfig
index 2997382d5..5e77ce08e 100644
--- a/net/packet/Kconfig
+++ b/net/packet/Kconfig
@@ -23,3 +23,13 @@ config PACKET_DIAG
help
Support for PF_PACKET sockets monitoring interface used by the ss tool.
If unsure, say Y.
+
+config PACKET_HRTIMER
+ tristate "Packet: use hrtimer instead of timer"
+ depends on PACKET
+ default n
+ help
+ Support hrtimer mode for PF_PACKET sockets to improve real-time
+ performance. The default timeout mechanism with jiffies granularity
+ is insufficient to meet the demands for real-time performance.
+ If unsure, say N.
diff --git a/net/packet/af_packet.c b/net/packet/af_packet.c
index bc438d0d9..5c7e4ef89 100644
--- a/net/packet/af_packet.c
+++ b/net/packet/af_packet.c
@@ -203,7 +203,11 @@ static void prb_retire_current_block(struct tpacket_kbdq_core *,
static int prb_queue_frozen(struct tpacket_kbdq_core *);
static void prb_open_block(struct tpacket_kbdq_core *,
struct tpacket_block_desc *);
+#ifdef CONFIG_PACKET_HRTIMER
+static enum hrtimer_restart prb_retire_rx_blk_timer_expired(struct hrtimer *);
+#else
static void prb_retire_rx_blk_timer_expired(struct timer_list *);
+#endif
static void _prb_refresh_rx_retire_blk_timer(struct tpacket_kbdq_core *);
static void prb_fill_rxhash(struct tpacket_kbdq_core *, struct tpacket3_hdr *);
static void prb_clear_rxhash(struct tpacket_kbdq_core *,
@@ -581,7 +585,11 @@ static __be16 vlan_get_protocol_dgram(const struct sk_buff *skb)
static void prb_del_retire_blk_timer(struct tpacket_kbdq_core *pkc)
{
+#ifdef CONFIG_PACKET_HRTIMER
+ hrtimer_cancel(&pkc->retire_blk_timer);
+#else
timer_delete_sync(&pkc->retire_blk_timer);
+#endif
}
static void prb_shutdown_retire_blk_timer(struct packet_sock *po,
@@ -603,9 +611,16 @@ static void prb_setup_retire_blk_timer(struct packet_sock *po)
struct tpacket_kbdq_core *pkc;
pkc = GET_PBDQC_FROM_RB(&po->rx_ring);
+#ifdef CONFIG_PACKET_HRTIMER
+ hrtimer_init(&pkc->retire_blk_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL_SOFT);
+ pkc->retire_blk_timer.function = prb_retire_rx_blk_timer_expired;
+ if (pkc->tov_in_msecs == 0)
+ pkc->tov_in_msecs = jiffies_to_msecs(1);
+#else
timer_setup(&pkc->retire_blk_timer, prb_retire_rx_blk_timer_expired,
0);
pkc->retire_blk_timer.expires = jiffies;
+#endif
}
static int prb_calc_retire_blk_tmo(struct packet_sock *po,
@@ -676,7 +691,11 @@ static void init_prb_bdqc(struct packet_sock *po,
else
p1->retire_blk_tov = prb_calc_retire_blk_tmo(po,
req_u->req3.tp_block_size);
+#ifdef CONFIG_PACKET_HRTIMER
+ p1->tov_in_msecs = p1->retire_blk_tov;
+#else
p1->tov_in_jiffies = msecs_to_jiffies(p1->retire_blk_tov);
+#endif
p1->blk_sizeof_priv = req_u->req3.tp_sizeof_priv;
rwlock_init(&p1->blk_fill_in_prog_lock);
@@ -691,8 +710,15 @@ static void init_prb_bdqc(struct packet_sock *po,
*/
static void _prb_refresh_rx_retire_blk_timer(struct tpacket_kbdq_core *pkc)
{
+#ifdef CONFIG_PACKET_HRTIMER
+ hrtimer_start_range_ns(&pkc->retire_blk_timer,
+ ms_to_ktime(pkc->tov_in_msecs),
+ 0,
+ HRTIMER_MODE_REL_SOFT);
+#else
mod_timer(&pkc->retire_blk_timer,
jiffies + pkc->tov_in_jiffies);
+#endif
pkc->last_kactive_blk_num = pkc->kactive_blk_num;
}
@@ -719,8 +745,15 @@ static void _prb_refresh_rx_retire_blk_timer(struct tpacket_kbdq_core *pkc)
* prb_calc_retire_blk_tmo() calculates the tmo.
*
*/
+#ifdef CONFIG_PACKET_HRTIMER
+static enum hrtimer_restart prb_retire_rx_blk_timer_expired(struct hrtimer *t)
+#else
static void prb_retire_rx_blk_timer_expired(struct timer_list *t)
+#endif
{
+#ifdef CONFIG_PACKET_HRTIMER
+ enum hrtimer_restart ret = HRTIMER_RESTART;
+#endif
struct packet_sock *po =
timer_container_of(po, t, rx_ring.prb_bdqc.retire_blk_timer);
struct tpacket_kbdq_core *pkc = GET_PBDQC_FROM_RB(&po->rx_ring);
@@ -787,9 +820,15 @@ static void prb_retire_rx_blk_timer_expired(struct timer_list *t)
refresh_timer:
_prb_refresh_rx_retire_blk_timer(pkc);
+#ifdef CONFIG_PACKET_HRTIMER
+ ret = HRTIMER_RESTART;
+#endif
out:
spin_unlock(&po->sk.sk_receive_queue.lock);
+#ifdef CONFIG_PACKET_HRTIMER
+ return ret;
+#endif
}
static void prb_flush_block(struct tpacket_kbdq_core *pkc1,
diff --git a/net/packet/internal.h b/net/packet/internal.h
index 1e743d031..d9f2b2492 100644
--- a/net/packet/internal.h
+++ b/net/packet/internal.h
@@ -47,10 +47,18 @@ struct tpacket_kbdq_core {
unsigned short retire_blk_tov;
unsigned short version;
+#ifdef CONFIG_PACKET_HRTIMER
+ unsigned long tov_in_msecs;
+#else
unsigned long tov_in_jiffies;
+#endif
/* timer to retire an outstanding block */
+#ifdef CONFIG_PACKET_HRTIMER
+ struct hrtimer retire_blk_timer;
+#else
struct timer_list retire_blk_timer;
+#endif
};
struct pgv {
--
2.34.1On Wed, 2025-08-06 at 13:52 +0800, Xin Zhao wrote: > In a system with high real-time requirements, the timeout mechanism of > ordinary timers with jiffies granularity is insufficient to meet the > demands for real-time performance. Meanwhile, the optimization of CPU > usage with af_packet is quite significant. Add hrtimer mode to help > compensate for the shortcomings in real-time performance. Do you have performance numbers? It would be nice to see the test environment, measurements carried out and some latency/jitter numbers. Thanks, Ferenc
On Tue, Aug 5, 2025 at 10:52 PM Xin Zhao <jackzxcui1989@163.com> wrote: > > In a system with high real-time requirements, the timeout mechanism of > ordinary timers with jiffies granularity is insufficient to meet the > demands for real-time performance. Meanwhile, the optimization of CPU > usage with af_packet is quite significant. Add hrtimer mode to help > compensate for the shortcomings in real-time performance. > > Signed-off-by: Xin Zhao <jackzxcui1989@163.com> I doubt we need another CONFIG option ? Also this seems to be beneficial for HZ=100 or HZ=250 configuration, maybe worth mentioning in the changelog. But more importantly, I think you should explain what difference this really makes, in which circumstances this timer needs to fire... If real-time is a concern, maybe using a timer to perform GC-style operation is a no go anyway...
Hi Xin,
kernel test robot noticed the following build errors:
[auto build test ERROR on net-next/main]
[also build test ERROR on net/main linus/master v6.16 next-20250806]
[cannot apply to horms-ipvs/master]
[If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]
url: https://github.com/intel-lab-lkp/linux/commits/Xin-Zhao/net-af_packet-add-af_packet-hrtimer-mode/20250806-135354
base: net-next/main
patch link: https://lore.kernel.org/r/20250806055210.1530081-1-jackzxcui1989%40163.com
patch subject: [PATCH] net: af_packet: add af_packet hrtimer mode
config: x86_64-randconfig-004-20250807 (https://download.01.org/0day-ci/archive/20250807/202508070419.1T0vxudH-lkp@intel.com/config)
compiler: gcc-12 (Debian 12.2.0-14+deb12u1) 12.2.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20250807/202508070419.1T0vxudH-lkp@intel.com/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202508070419.1T0vxudH-lkp@intel.com/
All errors (new ones prefixed by >>):
net/packet/af_packet.c: In function 'prb_setup_retire_blk_timer':
>> net/packet/af_packet.c:615:9: error: implicit declaration of function 'hrtimer_init'; did you mean 'hrtimers_init'? [-Werror=implicit-function-declaration]
615 | hrtimer_init(&pkc->retire_blk_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL_SOFT);
| ^~~~~~~~~~~~
| hrtimers_init
cc1: some warnings being treated as errors
vim +615 net/packet/af_packet.c
608
609 static void prb_setup_retire_blk_timer(struct packet_sock *po)
610 {
611 struct tpacket_kbdq_core *pkc;
612
613 pkc = GET_PBDQC_FROM_RB(&po->rx_ring);
614 #ifdef CONFIG_PACKET_HRTIMER
> 615 hrtimer_init(&pkc->retire_blk_timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL_SOFT);
616 pkc->retire_blk_timer.function = prb_retire_rx_blk_timer_expired;
617 if (pkc->tov_in_msecs == 0)
618 pkc->tov_in_msecs = jiffies_to_msecs(1);
619 #else
620 timer_setup(&pkc->retire_blk_timer, prb_retire_rx_blk_timer_expired,
621 0);
622 pkc->retire_blk_timer.expires = jiffies;
623 #endif
624 }
625
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
© 2016 - 2025 Red Hat, Inc.