[PATCH 0/4] workqueue: Some cleanups

[PATCH 0/4] workqueue: Some cleanups

[Lai Jiangshan]

From: Lai Jiangshan <jiangshan.ljs@antgroup.com>

Some simple cleanups.

Lai Jiangshan (4):
  workqueue: Use INIT_WORK_ONSTACK in workqueue_softirq_dead()
  workqueue: Reorder the fields in struct workqueue_attrs
  workqueue: Move attrs->cpumask out of worker_pool's properties when
    attrs->affn_strict
  workqueue: Use list_last_entry() to get the last idle worker

 include/linux/workqueue.h | 27 +++++++++++++++------------
 kernel/workqueue.c        | 19 +++++++++++--------
 2 files changed, 26 insertions(+), 20 deletions(-)

-- 
2.19.1.6.gb485710b

[PATCH 1/4] workqueue: Use INIT_WORK_ONSTACK in workqueue_softirq_dead()

[Lai Jiangshan]

From: Lai Jiangshan <jiangshan.ljs@antgroup.com>

dead_work is a stack variable.

Signed-off-by: Lai Jiangshan <jiangshan.ljs@antgroup.com>
---
 kernel/workqueue.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index a60eb65955e7..3ff91243322d 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -3687,7 +3687,7 @@ void workqueue_softirq_dead(unsigned int cpu)
 		if (!need_more_worker(pool))
 			continue;
 
-		INIT_WORK(&dead_work.work, drain_dead_softirq_workfn);
+		INIT_WORK_ONSTACK(&dead_work.work, drain_dead_softirq_workfn);
 		dead_work.pool = pool;
 		init_completion(&dead_work.done);
 
-- 
2.19.1.6.gb485710b

Re: [PATCH 1/4] workqueue: Use INIT_WORK_ONSTACK in workqueue_softirq_dead()

[Tejun Heo]

On Fri, Mar 08, 2024 at 05:42:50PM +0800, Lai Jiangshan wrote:
> From: Lai Jiangshan <jiangshan.ljs@antgroup.com>
> 
> dead_work is a stack variable.
> 
> Signed-off-by: Lai Jiangshan <jiangshan.ljs@antgroup.com>

Applied to wq/for-6.10.

Thanks.

-- 
tejun

[PATCH 2/4] workqueue: Reorder the fields in struct workqueue_attrs

[Lai Jiangshan]

From: Lai Jiangshan <jiangshan.ljs@antgroup.com>

So that its size is reduced from 40 to 32 in 64bit kernel, and it can be
saved more when allocated with kmalloc() in alloc_workqueue_attrs().

Signed-off-by: Lai Jiangshan <jiangshan.ljs@antgroup.com>
---
 include/linux/workqueue.h | 24 ++++++++++++------------
 1 file changed, 12 insertions(+), 12 deletions(-)

diff --git a/include/linux/workqueue.h b/include/linux/workqueue.h
index 158784dd189a..c885a5f6bb93 100644
--- a/include/linux/workqueue.h
+++ b/include/linux/workqueue.h
@@ -145,13 +145,15 @@ struct workqueue_attrs {
 	int nice;
 
 	/**
-	 * @cpumask: allowed CPUs
+	 * @affn_strict: affinity scope is strict
 	 *
-	 * Work items in this workqueue are affine to these CPUs and not allowed
-	 * to execute on other CPUs. A pool serving a workqueue must have the
-	 * same @cpumask.
+	 * If clear, workqueue will make a best-effort attempt at starting the
+	 * worker inside @__pod_cpumask but the scheduler is free to migrate it
+	 * outside.
+	 *
+	 * If set, workers are only allowed to run inside @__pod_cpumask.
 	 */
-	cpumask_var_t cpumask;
+	bool affn_strict;
 
 	/**
 	 * @__pod_cpumask: internal attribute used to create per-pod pools
@@ -166,15 +168,13 @@ struct workqueue_attrs {
 	cpumask_var_t __pod_cpumask;
 
 	/**
-	 * @affn_strict: affinity scope is strict
-	 *
-	 * If clear, workqueue will make a best-effort attempt at starting the
-	 * worker inside @__pod_cpumask but the scheduler is free to migrate it
-	 * outside.
+	 * @cpumask: allowed CPUs
 	 *
-	 * If set, workers are only allowed to run inside @__pod_cpumask.
+	 * Work items in this workqueue are affine to these CPUs and not allowed
+	 * to execute on other CPUs. A pool serving a workqueue must have the
+	 * same @cpumask.
 	 */
-	bool affn_strict;
+	cpumask_var_t cpumask;
 
 	/*
 	 * Below fields aren't properties of a worker_pool. They only modify how
-- 
2.19.1.6.gb485710b

Re: [PATCH 2/4] workqueue: Reorder the fields in struct workqueue_attrs

[Tejun Heo]

Hello,

On Fri, Mar 08, 2024 at 05:42:51PM +0800, Lai Jiangshan wrote:
> From: Lai Jiangshan <jiangshan.ljs@antgroup.com>
> 
> So that its size is reduced from 40 to 32 in 64bit kernel, and it can be
> saved more when allocated with kmalloc() in alloc_workqueue_attrs().

Does this matter tho? Also, the sizing would be dependent on the number of
supported CPUs, right?

So, before, the order was - nice, cpumasks, affinity scope related stuff,
and then ordered flag. After, the cpumasks and affinity scope stuff are
mixed. Unless the saving is actually meaningful, I'd rather keep the current
ordering.

Thanks.

-- 
tejun

[PATCH 3/4] workqueue: Move attrs->cpumask out of worker_pool's properties when attrs->affn_strict

[Lai Jiangshan]

From: Lai Jiangshan <jiangshan.ljs@antgroup.com>

Allow more pools can be shared when attrs->affn_strict.

Signed-off-by: Lai Jiangshan <jiangshan.ljs@antgroup.com>
---
 include/linux/workqueue.h |  3 +++
 kernel/workqueue.c        | 13 ++++++++-----
 2 files changed, 11 insertions(+), 5 deletions(-)

diff --git a/include/linux/workqueue.h b/include/linux/workqueue.h
index c885a5f6bb93..86f723579a32 100644
--- a/include/linux/workqueue.h
+++ b/include/linux/workqueue.h
@@ -180,6 +180,9 @@ struct workqueue_attrs {
 	 * Below fields aren't properties of a worker_pool. They only modify how
 	 * :c:func:`apply_workqueue_attrs` select pools and thus don't
 	 * participate in pool hash calculations or equality comparisons.
+	 *
+	 * If @affn_strict is set, @cpumask isn't a property of a worker_pool
+	 * either.
 	 */
 
 	/**
diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index 3ff91243322d..3bb71cfd3e06 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -4517,6 +4517,8 @@ static void wqattrs_clear_for_pool(struct workqueue_attrs *attrs)
 {
 	attrs->affn_scope = WQ_AFFN_NR_TYPES;
 	attrs->ordered = false;
+	if (attrs->affn_strict)
+		cpumask_copy(attrs->cpumask, cpu_possible_mask);
 }
 
 /* hash value of the content of @attr */
@@ -4525,11 +4527,12 @@ static u32 wqattrs_hash(const struct workqueue_attrs *attrs)
 	u32 hash = 0;
 
 	hash = jhash_1word(attrs->nice, hash);
-	hash = jhash(cpumask_bits(attrs->cpumask),
-		     BITS_TO_LONGS(nr_cpumask_bits) * sizeof(long), hash);
+	hash = jhash_1word(attrs->affn_strict, hash);
 	hash = jhash(cpumask_bits(attrs->__pod_cpumask),
 		     BITS_TO_LONGS(nr_cpumask_bits) * sizeof(long), hash);
-	hash = jhash_1word(attrs->affn_strict, hash);
+	if (!attrs->affn_strict)
+		hash = jhash(cpumask_bits(attrs->cpumask),
+			     BITS_TO_LONGS(nr_cpumask_bits) * sizeof(long), hash);
 	return hash;
 }
 
@@ -4539,11 +4542,11 @@ static bool wqattrs_equal(const struct workqueue_attrs *a,
 {
 	if (a->nice != b->nice)
 		return false;
-	if (!cpumask_equal(a->cpumask, b->cpumask))
+	if (a->affn_strict != b->affn_strict)
 		return false;
 	if (!cpumask_equal(a->__pod_cpumask, b->__pod_cpumask))
 		return false;
-	if (a->affn_strict != b->affn_strict)
+	if (!a->affn_strict && !cpumask_equal(a->cpumask, b->cpumask))
 		return false;
 	return true;
 }
-- 
2.19.1.6.gb485710b

Re: [PATCH 3/4] workqueue: Move attrs->cpumask out of worker_pool's properties when attrs->affn_strict

[Tejun Heo]

On Fri, Mar 08, 2024 at 05:42:52PM +0800, Lai Jiangshan wrote:
> From: Lai Jiangshan <jiangshan.ljs@antgroup.com>
> 
> Allow more pools can be shared when attrs->affn_strict.
> 
> Signed-off-by: Lai Jiangshan <jiangshan.ljs@antgroup.com>

Applied to wq/for-6.10.

Thanks.

-- 
tejun

[PATCH 4/4] workqueue: Use list_last_entry() to get the last idle worker

[Lai Jiangshan]

From: Lai Jiangshan <jiangshan.ljs@antgroup.com>

It is clearer than open code.

Signed-off-by: Lai Jiangshan <jiangshan.ljs@antgroup.com>
---
 kernel/workqueue.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index 3bb71cfd3e06..7a3392aab395 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -2940,7 +2940,7 @@ static void idle_worker_timeout(struct timer_list *t)
 		unsigned long expires;
 
 		/* idle_list is kept in LIFO order, check the last one */
-		worker = list_entry(pool->idle_list.prev, struct worker, entry);
+		worker = list_last_entry(&pool->idle_list, struct worker, entry);
 		expires = worker->last_active + IDLE_WORKER_TIMEOUT;
 		do_cull = !time_before(jiffies, expires);
 
@@ -2982,7 +2982,7 @@ static void idle_cull_fn(struct work_struct *work)
 		struct worker *worker;
 		unsigned long expires;
 
-		worker = list_entry(pool->idle_list.prev, struct worker, entry);
+		worker = list_last_entry(&pool->idle_list, struct worker, entry);
 		expires = worker->last_active + IDLE_WORKER_TIMEOUT;
 
 		if (time_before(jiffies, expires)) {
-- 
2.19.1.6.gb485710b

Re: [PATCH 4/4] workqueue: Use list_last_entry() to get the last idle worker

[Tejun Heo]

On Fri, Mar 08, 2024 at 05:42:53PM +0800, Lai Jiangshan wrote:
> From: Lai Jiangshan <jiangshan.ljs@antgroup.com>
> 
> It is clearer than open code.
> 
> Signed-off-by: Lai Jiangshan <jiangshan.ljs@antgroup.com>

Applied to wq/for-6.10.

Thanks.

-- 
tejun