[PATCH 0/3] replace old wq(s), added WQ_PERCPU to alloc_workqueue

[PATCH 0/3] replace old wq(s), added WQ_PERCPU to alloc_workqueue

[Marco Crivellari]

Hi,

=== Current situation: problems ===

Let's consider a nohz_full system with isolated CPUs: wq_unbound_cpumask is
set to the housekeeping CPUs, for !WQ_UNBOUND the local CPU is selected.

This leads to different scenarios if a work item is scheduled on an
isolated CPU where "delay" value is 0 or greater then 0:
        schedule_delayed_work(, 0);

This will be handled by __queue_work() that will queue the work item on the
current local (isolated) CPU, while:

        schedule_delayed_work(, 1);

Will move the timer on an housekeeping CPU, and schedule the work there.

Currently if a user enqueue a work item using schedule_delayed_work() the
used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use
WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to
schedule_work() that is using system_wq and queue_work(), that makes use
again of WORK_CPU_UNBOUND.

This lack of consistency cannot be addressed without refactoring the API.

=== Recent changes to the WQ API ===

The following, address the recent changes in the Workqueue API:

- commit 128ea9f6ccfb ("workqueue: Add system_percpu_wq and system_dfl_wq")
- commit 930c2ea566af ("workqueue: Add new WQ_PERCPU flag")

The old workqueues will be removed in a future release cycle.

=== Introduced Changes by this series ===

1) [P 1] Replace uses of system_wq

    system_wq is a per-CPU workqueue, but his name is not clear.
    Because of that, system_wq has been replaced with system_percpu_wq.

2) [P 2-3] add WQ_PERCPU to all relevant alloc_workqueue() users

    This change adds a new WQ_PERCPU flag to explicitly request
    alloc_workqueue() to be per-cpu when WQ_UNBOUND has not been specified.


Thanks!

Marco Crivellari (3):
  nvmet: replace use of system_wq with system_percpu_wq
  nvme: add WQ_PERCPU to alloc_workqueue users
  nvmet-fc: add WQ_PERCPU to alloc_workqueue users

 drivers/nvme/target/admin-cmd.c        | 2 +-
 drivers/nvme/target/core.c             | 5 +++--
 drivers/nvme/target/fabrics-cmd-auth.c | 2 +-
 drivers/nvme/target/fc.c               | 6 +++---
 drivers/nvme/target/tcp.c              | 2 +-
 5 files changed, 9 insertions(+), 8 deletions(-)

-- 
2.51.1

[PATCH 1/3] nvmet: replace use of system_wq with system_percpu_wq

[Marco Crivellari]

Currently if a user enqueue a work item using schedule_delayed_work() the
used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use
WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to
schedule_work() that is using system_wq and queue_work(), that makes use
again of WORK_CPU_UNBOUND.

This lack of consistentcy cannot be addressed without refactoring the API.

This patch continues the effort to refactor worqueue APIs, which has begun
with the change introducing new workqueues and a new alloc_workqueue flag:

commit 128ea9f6ccfb ("workqueue: Add system_percpu_wq and system_dfl_wq")
commit 930c2ea566af ("workqueue: Add new WQ_PERCPU flag")

system_wq should be the per-cpu workqueue, yet in this name nothing makes
that clear, so replace system_wq with system_percpu_wq.

The old wq (system_wq) will be kept for a few release cycles.

Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
---
 drivers/nvme/target/admin-cmd.c        | 2 +-
 drivers/nvme/target/fabrics-cmd-auth.c | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/nvme/target/admin-cmd.c b/drivers/nvme/target/admin-cmd.c
index 3e378153a781..29281f410cc6 100644
--- a/drivers/nvme/target/admin-cmd.c
+++ b/drivers/nvme/target/admin-cmd.c
@@ -1604,7 +1604,7 @@ void nvmet_execute_keep_alive(struct nvmet_req *req)
 
 	pr_debug("ctrl %d update keep-alive timer for %d secs\n",
 		ctrl->cntlid, ctrl->kato);
-	mod_delayed_work(system_wq, &ctrl->ka_work, ctrl->kato * HZ);
+	mod_delayed_work(system_percpu_wq, &ctrl->ka_work, ctrl->kato * HZ);
 out:
 	nvmet_req_complete(req, status);
 }
diff --git a/drivers/nvme/target/fabrics-cmd-auth.c b/drivers/nvme/target/fabrics-cmd-auth.c
index bf01ec414c55..8f504bf891de 100644
--- a/drivers/nvme/target/fabrics-cmd-auth.c
+++ b/drivers/nvme/target/fabrics-cmd-auth.c
@@ -390,7 +390,7 @@ void nvmet_execute_auth_send(struct nvmet_req *req)
 	    req->sq->dhchap_step != NVME_AUTH_DHCHAP_MESSAGE_FAILURE2) {
 		unsigned long auth_expire_secs = ctrl->kato ? ctrl->kato : 120;
 
-		mod_delayed_work(system_wq, &req->sq->auth_expired_work,
+		mod_delayed_work(system_percpu_wq, &req->sq->auth_expired_work,
 				 auth_expire_secs * HZ);
 		goto complete;
 	}
-- 
2.51.1

[PATCH 2/3] nvme: add WQ_PERCPU to alloc_workqueue users

[Marco Crivellari]

Currently if a user enqueue a work item using schedule_delayed_work() the
used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use
WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to
schedule_work() that is using system_wq and queue_work(), that makes use
again of WORK_CPU_UNBOUND.
This lack of consistentcy cannot be addressed without refactoring the API.

alloc_workqueue() treats all queues as per-CPU by default, while unbound
workqueues must opt-in via WQ_UNBOUND.

This default is suboptimal: most workloads benefit from unbound queues,
allowing the scheduler to place worker threads where they’re needed and
reducing noise when CPUs are isolated.

This change adds a new WQ_PERCPU flag to explicitly request alloc_workqueue()
to be per-cpu when WQ_UNBOUND has not been specified.

This patch continues the effort to refactor worqueue APIs, which has begun
with the change introducing new workqueues and a new alloc_workqueue flag:

commit 128ea9f6ccfb ("workqueue: Add system_percpu_wq and system_dfl_wq")
commit 930c2ea566af ("workqueue: Add new WQ_PERCPU flag")

With the introduction of the WQ_PERCPU flag (equivalent to !WQ_UNBOUND),
any alloc_workqueue() caller that doesn’t explicitly specify WQ_UNBOUND
must now use WQ_PERCPU.

Once migration is complete, WQ_UNBOUND can be removed and unbound will
become the implicit default.

Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
---
 drivers/nvme/target/core.c | 5 +++--
 drivers/nvme/target/tcp.c  | 2 +-
 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/drivers/nvme/target/core.c b/drivers/nvme/target/core.c
index 5d7d483bfbe3..ae8e903b55ad 100644
--- a/drivers/nvme/target/core.c
+++ b/drivers/nvme/target/core.c
@@ -1943,12 +1943,13 @@ static int __init nvmet_init(void)
 	if (!nvmet_bvec_cache)
 		return -ENOMEM;
 
-	zbd_wq = alloc_workqueue("nvmet-zbd-wq", WQ_MEM_RECLAIM, 0);
+	zbd_wq = alloc_workqueue("nvmet-zbd-wq", WQ_MEM_RECLAIM | WQ_PERCPU,
+				 0);
 	if (!zbd_wq)
 		goto out_destroy_bvec_cache;
 
 	buffered_io_wq = alloc_workqueue("nvmet-buffered-io-wq",
-			WQ_MEM_RECLAIM, 0);
+			WQ_MEM_RECLAIM | WQ_PERCPU, 0);
 	if (!buffered_io_wq)
 		goto out_free_zbd_work_queue;
 
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index 470bf37e5a63..fa7a56626bcc 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -2198,7 +2198,7 @@ static int __init nvmet_tcp_init(void)
 	int ret;
 
 	nvmet_tcp_wq = alloc_workqueue("nvmet_tcp_wq",
-				WQ_MEM_RECLAIM | WQ_HIGHPRI, 0);
+				WQ_MEM_RECLAIM | WQ_HIGHPRI | WQ_PERCPU, 0);
 	if (!nvmet_tcp_wq)
 		return -ENOMEM;
 
-- 
2.51.1

[PATCH 3/3] nvmet-fc: add WQ_PERCPU to alloc_workqueue users

[Marco Crivellari]

Currently if a user enqueue a work item using schedule_delayed_work() the
used wq is "system_wq" (per-cpu wq) while queue_delayed_work() use
WORK_CPU_UNBOUND (used when a cpu is not specified). The same applies to
schedule_work() that is using system_wq and queue_work(), that makes use
again of WORK_CPU_UNBOUND.
This lack of consistentcy cannot be addressed without refactoring the API.

alloc_workqueue() treats all queues as per-CPU by default, while unbound
workqueues must opt-in via WQ_UNBOUND.

This default is suboptimal: most workloads benefit from unbound queues,
allowing the scheduler to place worker threads where they’re needed and
reducing noise when CPUs are isolated.

This change adds a new WQ_PERCPU flag to explicitly request alloc_workqueue()
to be per-cpu when WQ_UNBOUND has not been specified.

This patch continues the effort to refactor worqueue APIs, which has begun
with the change introducing new workqueues and a new alloc_workqueue flag:

commit 128ea9f6ccfb ("workqueue: Add system_percpu_wq and system_dfl_wq")
commit 930c2ea566af ("workqueue: Add new WQ_PERCPU flag")

With the introduction of the WQ_PERCPU flag (equivalent to !WQ_UNBOUND),
any alloc_workqueue() caller that doesn’t explicitly specify WQ_UNBOUND
must now use WQ_PERCPU.

Once migration is complete, WQ_UNBOUND can be removed and unbound will
become the implicit default.

Cc: Justin Tee <justin.tee@broadcom.com>
Cc: Naresh Gottumukkala <nareshgottumukkala83@gmail.com>
CC: Paul Ely <paul.ely@broadcom.com>
Suggested-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Marco Crivellari <marco.crivellari@suse.com>
---
 drivers/nvme/target/fc.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/nvme/target/fc.c b/drivers/nvme/target/fc.c
index 7d84527d5a43..ac730f2e6ea1 100644
--- a/drivers/nvme/target/fc.c
+++ b/drivers/nvme/target/fc.c
@@ -795,9 +795,9 @@ nvmet_fc_alloc_target_queue(struct nvmet_fc_tgt_assoc *assoc,
 	if (!queue)
 		return NULL;
 
-	queue->work_q = alloc_workqueue("ntfc%d.%d.%d", 0, 0,
-				assoc->tgtport->fc_target_port.port_num,
-				assoc->a_id, qid);
+	queue->work_q = alloc_workqueue("ntfc%d.%d.%d", WQ_PERCPU, 0,
+					assoc->tgtport->fc_target_port.port_num,
+					assoc->a_id, qid);
 	if (!queue->work_q)
 		goto out_free_queue;
 
-- 
2.51.1

Re: [PATCH 3/3] nvmet-fc: add WQ_PERCPU to alloc_workqueue users

[Justin Tee]

Reviewed-by: Justin Tee <justin.tee@broadcom.com>

Regards,
Justin

Re: [PATCH 3/3] nvmet-fc: add WQ_PERCPU to alloc_workqueue users

[Marco Crivellari]

On Tue, Nov 4, 2025 at 12:25 PM Marco Crivellari
<marco.crivellari@suse.com> wrote:
> [...]
>  drivers/nvme/target/fc.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/nvme/target/fc.c b/drivers/nvme/target/fc.c
> index 7d84527d5a43..ac730f2e6ea1 100644
> --- a/drivers/nvme/target/fc.c
> +++ b/drivers/nvme/target/fc.c

Hi,

Gentle ping.

Thanks!

-- 

Marco Crivellari

L3 Support Engineer