[RFC PATCH net-next] net: bridge: Move long delayed work on system_dfl_long_wq

[RFC PATCH net-next] net: bridge: Move long delayed work on system_dfl_long_wq

[Marco Crivellari]

Currently the code enqueue work items using {queue|mod}_delayed_work(),
using system_long_wq. This workqueue should be used when long works are
expected and it is a per-cpu workqueue.

The function(s) end up calling __queue_delayed_work(), which set a global
timer that could fire anywhere, enqueuing the work where the timer fired.

Unbound works could benefit from scheduler task placement, to optimize
performance and power consumption. Long work shouldn't stick to a single
CPU.

Recently, a new unbound workqueue specific for long running work has
been added:

    c116737e972e ("workqueue: Add system_dfl_long_wq for long unbound works")

Since the workqueue work doesn't rely on per-cpu variables, there is no
obvious reason that justify the use of a per-cpu workqueue. So change
system_long_wq with system_dfl_long_wq so that the work may benefit from
scheduler task placement.

Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
---
 net/bridge/br_fdb.c    | 2 +-
 net/bridge/br_stp.c    | 2 +-
 net/bridge/br_stp_if.c | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/net/bridge/br_fdb.c b/net/bridge/br_fdb.c
index ac81e58d5f70..14c7943d59a1 100644
--- a/net/bridge/br_fdb.c
+++ b/net/bridge/br_fdb.c
@@ -583,7 +583,7 @@ void br_fdb_cleanup(struct work_struct *work)
 
 	/* Cleanup minimum 10 milliseconds apart */
 	work_delay = max_t(unsigned long, work_delay, msecs_to_jiffies(10));
-	mod_delayed_work(system_long_wq, &br->gc_work, work_delay);
+	mod_delayed_work(system_dfl_long_wq, &br->gc_work, work_delay);
 }
 
 static void br_fdb_delete_locals_per_vlan_port(struct net_bridge *br,
diff --git a/net/bridge/br_stp.c b/net/bridge/br_stp.c
index 024210f95468..426e1c76ef6e 100644
--- a/net/bridge/br_stp.c
+++ b/net/bridge/br_stp.c
@@ -640,7 +640,7 @@ int br_set_ageing_time(struct net_bridge *br, clock_t ageing_time)
 	br->ageing_time = t;
 	spin_unlock_bh(&br->lock);
 
-	mod_delayed_work(system_long_wq, &br->gc_work, 0);
+	mod_delayed_work(system_dfl_long_wq, &br->gc_work, 0);
 
 	return 0;
 }
diff --git a/net/bridge/br_stp_if.c b/net/bridge/br_stp_if.c
index 28c1d3f7e22f..7a5f4747e1a1 100644
--- a/net/bridge/br_stp_if.c
+++ b/net/bridge/br_stp_if.c
@@ -53,7 +53,7 @@ void br_stp_enable_bridge(struct net_bridge *br)
 	spin_lock_bh(&br->lock);
 	if (br->stp_enabled == BR_KERNEL_STP)
 		mod_timer(&br->hello_timer, jiffies + br->hello_time);
-	mod_delayed_work(system_long_wq, &br->gc_work, HZ / 10);
+	mod_delayed_work(system_dfl_long_wq, &br->gc_work, HZ / 10);
 
 	br_config_bpdu_generation(br);
 
-- 
2.53.0

Re: [RFC PATCH net-next] net: bridge: Move long delayed work on system_dfl_long_wq

[Ido Schimmel]

On Thu, May 07, 2026 at 03:48:15PM +0200, Marco Crivellari wrote:
> Currently the code enqueue work items using {queue|mod}_delayed_work(),
> using system_long_wq. This workqueue should be used when long works are
> expected and it is a per-cpu workqueue.
> 
> The function(s) end up calling __queue_delayed_work(), which set a global
> timer that could fire anywhere, enqueuing the work where the timer fired.
> 
> Unbound works could benefit from scheduler task placement, to optimize
> performance and power consumption. Long work shouldn't stick to a single
> CPU.
> 
> Recently, a new unbound workqueue specific for long running work has
> been added:
> 
>     c116737e972e ("workqueue: Add system_dfl_long_wq for long unbound works")
> 
> Since the workqueue work doesn't rely on per-cpu variables, there is no
> obvious reason that justify the use of a per-cpu workqueue. So change
> system_long_wq with system_dfl_long_wq so that the work may benefit from
> scheduler task placement.
> 
> Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>

Reviewed-by: Ido Schimmel <idosch@nvidia.com>