Merge tag 'for-6.10/dm-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm

Pull device mapper updates from Mike Snitzer:

 - Add a dm-crypt optional "high_priority" flag that enables the crypt
   workqueues to use WQ_HIGHPRI.

 - Export dm-crypt workqueues via sysfs (by enabling WQ_SYSFS) to allow
   for improved visibility and controls over IO and crypt workqueues.

 - Fix dm-crypt to no longer constrain max_segment_size to PAGE_SIZE.
   This limit isn't needed given that the block core provides late bio
   splitting if bio exceeds underlying limits (e.g. max_segment_size).

 - Fix dm-crypt crypt_queue's use of WQ_UNBOUND to not use
   WQ_CPU_INTENSIVE because it is meaningless with WQ_UNBOUND.

 - Fix various issues with dm-delay target (ranging from a resource
   teardown fix, a fix for hung task when using kthread mode, and other
   improvements that followed from code inspection).

* tag 'for-6.10/dm-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm-delay: remove timer_lock
  dm-delay: change locking to avoid contention
  dm-delay: fix max_delay calculations
  dm-delay: fix hung task introduced by kthread mode
  dm-delay: fix workqueue delay_timer race
  dm-crypt: don't set WQ_CPU_INTENSIVE for WQ_UNBOUND crypt_queue
  dm: use queue_limits_set
  dm-crypt: stop constraining max_segment_size to PAGE_SIZE
  dm-crypt: export sysfs of all workqueues
  dm-crypt: add the optional "high_priority" flag

4 files changed, 97 insertions, 68 deletions

diff --git a/Documentation/admin-guide/device-mapper/dm-crypt.rst b/Documentation/admin-guide/device-mapper/dm-crypt.rst
index aa2d04d95df6..41f5f57f00eb 100644
--- a/Documentation/admin-guide/device-mapper/dm-crypt.rst
+++ b/Documentation/admin-guide/device-mapper/dm-crypt.rst
@@ -113,6 +113,11 @@ same_cpu_crypt
     The default is to use an unbound workqueue so that encryption work
     is automatically balanced between available CPUs.
 
+high_priority
+    Set dm-crypt workqueues and the writer thread to high priority. This
+    improves throughput and latency of dm-crypt while degrading general
+    responsiveness of the system.
+
 submit_from_crypt_cpus
     Disable offloading writes to a separate thread after encryption.
     There are some situations where offloading write bios from the
diff --git a/drivers/md/dm-crypt.c b/drivers/md/dm-crypt.c
index 9a74c6316c5d..1b7a97cc3779 100644
--- a/drivers/md/dm-crypt.c
+++ b/drivers/md/dm-crypt.c
@@ -47,6 +47,8 @@
 
 #define DM_MSG_PREFIX "crypt"
 
+static DEFINE_IDA(workqueue_ida);
+
 /*
  * context holding the current state of a multi-part conversion
  */
@@ -137,9 +139,9 @@ struct iv_elephant_private {
  * and encrypts / decrypts at the same time.
  */
 enum flags { DM_CRYPT_SUSPENDED, DM_CRYPT_KEY_VALID,
-	     DM_CRYPT_SAME_CPU, DM_CRYPT_NO_OFFLOAD,
-	     DM_CRYPT_NO_READ_WORKQUEUE, DM_CRYPT_NO_WRITE_WORKQUEUE,
-	     DM_CRYPT_WRITE_INLINE };
+	     DM_CRYPT_SAME_CPU, DM_CRYPT_HIGH_PRIORITY,
+	     DM_CRYPT_NO_OFFLOAD, DM_CRYPT_NO_READ_WORKQUEUE,
+	     DM_CRYPT_NO_WRITE_WORKQUEUE, DM_CRYPT_WRITE_INLINE };
 
 enum cipher_flags {
 	CRYPT_MODE_INTEGRITY_AEAD,	/* Use authenticated mode for cipher */
@@ -184,6 +186,7 @@ struct crypt_config {
 		struct crypto_aead **tfms_aead;
 	} cipher_tfm;
 	unsigned int tfms_count;
+	int workqueue_id;
 	unsigned long cipher_flags;
 
 	/*
@@ -1653,8 +1656,8 @@ static void crypt_free_buffer_pages(struct crypt_config *cc, struct bio *clone);
 
 /*
  * Generate a new unfragmented bio with the given size
- * This should never violate the device limitations (but only because
- * max_segment_size is being constrained to PAGE_SIZE).
+ * This should never violate the device limitations (but if it did then block
+ * core should split the bio as needed).
  *
  * This function may be called concurrently. If we allocate from the mempool
  * concurrently, there is a possibility of deadlock. For example, if we have
@@ -2771,6 +2774,9 @@ static void crypt_dtr(struct dm_target *ti)
 	if (cc->crypt_queue)
 		destroy_workqueue(cc->crypt_queue);
 
+	if (cc->workqueue_id)
+		ida_free(&workqueue_ida, cc->workqueue_id);
+
 	crypt_free_tfms(cc);
 
 	bioset_exit(&cc->bs);
@@ -3134,7 +3140,7 @@ static int crypt_ctr_optional(struct dm_target *ti, unsigned int argc, char **ar
 	struct crypt_config *cc = ti->private;
 	struct dm_arg_set as;
 	static const struct dm_arg _args[] = {
-		{0, 8, "Invalid number of feature args"},
+		{0, 9, "Invalid number of feature args"},
 	};
 	unsigned int opt_params, val;
 	const char *opt_string, *sval;
@@ -3161,6 +3167,8 @@ static int crypt_ctr_optional(struct dm_target *ti, unsigned int argc, char **ar
 
 		else if (!strcasecmp(opt_string, "same_cpu_crypt"))
 			set_bit(DM_CRYPT_SAME_CPU, &cc->flags);
+		else if (!strcasecmp(opt_string, "high_priority"))
+			set_bit(DM_CRYPT_HIGH_PRIORITY, &cc->flags);
 
 		else if (!strcasecmp(opt_string, "submit_from_crypt_cpus"))
 			set_bit(DM_CRYPT_NO_OFFLOAD, &cc->flags);
@@ -3230,8 +3238,9 @@ static int crypt_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 {
 	struct crypt_config *cc;
 	const char *devname = dm_table_device_name(ti->table);
-	int key_size;
+	int key_size, wq_id;
 	unsigned int align_mask;
+	unsigned int common_wq_flags;
 	unsigned long long tmpll;
 	int ret;
 	size_t iv_size_padding, additional_req_size;
@@ -3398,20 +3407,38 @@ static int crypt_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 		cc->tag_pool_max_sectors <<= cc->sector_shift;
 	}
 
+	wq_id = ida_alloc_min(&workqueue_ida, 1, GFP_KERNEL);
+	if (wq_id < 0) {
+		ti->error = "Couldn't get workqueue id";
+		ret = wq_id;
+		goto bad;
+	}
+	cc->workqueue_id = wq_id;
+
 	ret = -ENOMEM;
-	cc->io_queue = alloc_workqueue("kcryptd_io/%s", WQ_MEM_RECLAIM, 1, devname);
+	common_wq_flags = WQ_MEM_RECLAIM | WQ_SYSFS;
+	if (test_bit(DM_CRYPT_HIGH_PRIORITY, &cc->flags))
+		common_wq_flags |= WQ_HIGHPRI;
+
+	cc->io_queue = alloc_workqueue("kcryptd_io-%s-%d", common_wq_flags, 1, devname, wq_id);
 	if (!cc->io_queue) {
 		ti->error = "Couldn't create kcryptd io queue";
 		goto bad;
 	}
 
-	if (test_bit(DM_CRYPT_SAME_CPU, &cc->flags))
-		cc->crypt_queue = alloc_workqueue("kcryptd/%s", WQ_CPU_INTENSIVE | WQ_MEM_RECLAIM,
-						  1, devname);
-	else
-		cc->crypt_queue = alloc_workqueue("kcryptd/%s",
-						  WQ_CPU_INTENSIVE | WQ_MEM_RECLAIM | WQ_UNBOUND,
-						  num_online_cpus(), devname);
+	if (test_bit(DM_CRYPT_SAME_CPU, &cc->flags)) {
+		cc->crypt_queue = alloc_workqueue("kcryptd-%s-%d",
+						  common_wq_flags | WQ_CPU_INTENSIVE,
+						  1, devname, wq_id);
+	} else {
+		/*
+		 * While crypt_queue is certainly CPU intensive, the use of
+		 * WQ_CPU_INTENSIVE is meaningless with WQ_UNBOUND.
+		 */
+		cc->crypt_queue = alloc_workqueue("kcryptd-%s-%d",
+						  common_wq_flags | WQ_UNBOUND,
+						  num_online_cpus(), devname, wq_id);
+	}
 	if (!cc->crypt_queue) {
 		ti->error = "Couldn't create kcryptd queue";
 		goto bad;
@@ -3427,6 +3454,8 @@ static int crypt_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 		ti->error = "Couldn't spawn write thread";
 		goto bad;
 	}
+	if (test_bit(DM_CRYPT_HIGH_PRIORITY, &cc->flags))
+		set_user_nice(cc->write_thread, MIN_NICE);
 
 	ti->num_flush_bios = 1;
 	ti->limit_swap_bios = true;
@@ -3547,6 +3576,7 @@ static void crypt_status(struct dm_target *ti, status_type_t type,
 
 		num_feature_args += !!ti->num_discard_bios;
 		num_feature_args += test_bit(DM_CRYPT_SAME_CPU, &cc->flags);
+		num_feature_args += test_bit(DM_CRYPT_HIGH_PRIORITY, &cc->flags);
 		num_feature_args += test_bit(DM_CRYPT_NO_OFFLOAD, &cc->flags);
 		num_feature_args += test_bit(DM_CRYPT_NO_READ_WORKQUEUE, &cc->flags);
 		num_feature_args += test_bit(DM_CRYPT_NO_WRITE_WORKQUEUE, &cc->flags);
@@ -3560,6 +3590,8 @@ static void crypt_status(struct dm_target *ti, status_type_t type,
 				DMEMIT(" allow_discards");
 			if (test_bit(DM_CRYPT_SAME_CPU, &cc->flags))
 				DMEMIT(" same_cpu_crypt");
+			if (test_bit(DM_CRYPT_HIGH_PRIORITY, &cc->flags))
+				DMEMIT(" high_priority");
 			if (test_bit(DM_CRYPT_NO_OFFLOAD, &cc->flags))
 				DMEMIT(" submit_from_crypt_cpus");
 			if (test_bit(DM_CRYPT_NO_READ_WORKQUEUE, &cc->flags))
@@ -3579,6 +3611,7 @@ static void crypt_status(struct dm_target *ti, status_type_t type,
 		DMEMIT_TARGET_NAME_VERSION(ti->type);
 		DMEMIT(",allow_discards=%c", ti->num_discard_bios ? 'y' : 'n');
 		DMEMIT(",same_cpu_crypt=%c", test_bit(DM_CRYPT_SAME_CPU, &cc->flags) ? 'y' : 'n');
+		DMEMIT(",high_priority=%c", test_bit(DM_CRYPT_HIGH_PRIORITY, &cc->flags) ? 'y' : 'n');
 		DMEMIT(",submit_from_crypt_cpus=%c", test_bit(DM_CRYPT_NO_OFFLOAD, &cc->flags) ?
 		       'y' : 'n');
 		DMEMIT(",no_read_workqueue=%c", test_bit(DM_CRYPT_NO_READ_WORKQUEUE, &cc->flags) ?
@@ -3688,14 +3721,6 @@ static void crypt_io_hints(struct dm_target *ti, struct queue_limits *limits)
 {
 	struct crypt_config *cc = ti->private;
 
-	/*
-	 * Unfortunate constraint that is required to avoid the potential
-	 * for exceeding underlying device's max_segments limits -- due to
-	 * crypt_alloc_buffer() possibly allocating pages for the encryption
-	 * bio that are not as physically contiguous as the original bio.
-	 */
-	limits->max_segment_size = PAGE_SIZE;
-
 	limits->logical_block_size =
 		max_t(unsigned int, limits->logical_block_size, cc->sector_size);
 	limits->physical_block_size =
@@ -3706,7 +3731,7 @@ static void crypt_io_hints(struct dm_target *ti, struct queue_limits *limits)
 
 static struct target_type crypt_target = {
 	.name   = "crypt",
-	.version = {1, 25, 0},
+	.version = {1, 26, 0},
 	.module = THIS_MODULE,
 	.ctr    = crypt_ctr,
 	.dtr    = crypt_dtr,
diff --git a/drivers/md/dm-delay.c b/drivers/md/dm-delay.c
index 5eabdb06c649..08f6387620c1 100644
--- a/drivers/md/dm-delay.c
+++ b/drivers/md/dm-delay.c
@@ -28,7 +28,8 @@ struct delay_class {
 
 struct delay_c {
 	struct timer_list delay_timer;
-	struct mutex timer_lock;
+	struct mutex process_bios_lock; /* hold while removing bios to be processed from list */
+	spinlock_t delayed_bios_lock; /* hold on all accesses to delayed_bios list */
 	struct workqueue_struct *kdelayd_wq;
 	struct work_struct flush_expired_bios;
 	struct list_head delayed_bios;
@@ -49,8 +50,6 @@ struct dm_delay_info {
 	unsigned long expires;
 };
 
-static DEFINE_MUTEX(delayed_bios_lock);
-
 static void handle_delayed_timer(struct timer_list *t)
 {
 	struct delay_c *dc = from_timer(dc, t, delay_timer);
@@ -60,12 +59,7 @@ static void handle_delayed_timer(struct timer_list *t)
 
 static void queue_timeout(struct delay_c *dc, unsigned long expires)
 {
-	mutex_lock(&dc->timer_lock);
-
-	if (!timer_pending(&dc->delay_timer) || expires < dc->delay_timer.expires)
-		mod_timer(&dc->delay_timer, expires);
-
-	mutex_unlock(&dc->timer_lock);
+	timer_reduce(&dc->delay_timer, expires);
 }
 
 static inline bool delay_is_fast(struct delay_c *dc)
@@ -89,12 +83,16 @@ static void flush_delayed_bios(struct delay_c *dc, bool flush_all)
 {
 	struct dm_delay_info *delayed, *next;
 	struct bio_list flush_bio_list;
+	LIST_HEAD(local_list);
 	unsigned long next_expires = 0;
 	bool start_timer = false;
 	bio_list_init(&flush_bio_list);
 
-	mutex_lock(&delayed_bios_lock);
-	list_for_each_entry_safe(delayed, next, &dc->delayed_bios, list) {
+	mutex_lock(&dc->process_bios_lock);
+	spin_lock(&dc->delayed_bios_lock);
+	list_replace_init(&dc->delayed_bios, &local_list);
+	spin_unlock(&dc->delayed_bios_lock);
+	list_for_each_entry_safe(delayed, next, &local_list, list) {
 		cond_resched();
 		if (flush_all || time_after_eq(jiffies, delayed->expires)) {
 			struct bio *bio = dm_bio_from_per_bio_data(delayed,
@@ -114,7 +112,10 @@ static void flush_delayed_bios(struct delay_c *dc, bool flush_all)
 			}
 		}
 	}
-	mutex_unlock(&delayed_bios_lock);
+	spin_lock(&dc->delayed_bios_lock);
+	list_splice(&local_list, &dc->delayed_bios);
+	spin_unlock(&dc->delayed_bios_lock);
+	mutex_unlock(&dc->process_bios_lock);
 
 	if (start_timer)
 		queue_timeout(dc, next_expires);
@@ -128,13 +129,13 @@ static int flush_worker_fn(void *data)
 
 	while (!kthread_should_stop()) {
 		flush_delayed_bios(dc, false);
-		mutex_lock(&delayed_bios_lock);
+		spin_lock(&dc->delayed_bios_lock);
 		if (unlikely(list_empty(&dc->delayed_bios))) {
 			set_current_state(TASK_INTERRUPTIBLE);
-			mutex_unlock(&delayed_bios_lock);
+			spin_unlock(&dc->delayed_bios_lock);
 			schedule();
 		} else {
-			mutex_unlock(&delayed_bios_lock);
+			spin_unlock(&dc->delayed_bios_lock);
 			cond_resched();
 		}
 	}
@@ -154,8 +155,10 @@ static void delay_dtr(struct dm_target *ti)
 {
 	struct delay_c *dc = ti->private;
 
-	if (dc->kdelayd_wq)
+	if (dc->kdelayd_wq) {
+		timer_shutdown_sync(&dc->delay_timer);
 		destroy_workqueue(dc->kdelayd_wq);
+	}
 
 	if (dc->read.dev)
 		dm_put_device(ti, dc->read.dev);
@@ -166,7 +169,7 @@ static void delay_dtr(struct dm_target *ti)
 	if (dc->worker)
 		kthread_stop(dc->worker);
 
-	mutex_destroy(&dc->timer_lock);
+	mutex_destroy(&dc->process_bios_lock);
 
 	kfree(dc);
 }
@@ -224,7 +227,8 @@ static int delay_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 
 	ti->private = dc;
 	INIT_LIST_HEAD(&dc->delayed_bios);
-	mutex_init(&dc->timer_lock);
+	mutex_init(&dc->process_bios_lock);
+	spin_lock_init(&dc->delayed_bios_lock);
 	dc->may_delay = true;
 	dc->argc = argc;
 
@@ -240,19 +244,18 @@ static int delay_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 		ret = delay_class_ctr(ti, &dc->flush, argv);
 		if (ret)
 			goto bad;
-		max_delay = max(max_delay, dc->write.delay);
-		max_delay = max(max_delay, dc->flush.delay);
 		goto out;
 	}
 
 	ret = delay_class_ctr(ti, &dc->write, argv + 3);
 	if (ret)
 		goto bad;
+	max_delay = max(max_delay, dc->write.delay);
+
 	if (argc == 6) {
 		ret = delay_class_ctr(ti, &dc->flush, argv + 3);
 		if (ret)
 			goto bad;
-		max_delay = max(max_delay, dc->flush.delay);
 		goto out;
 	}
 
@@ -267,8 +270,7 @@ out:
 		 * In case of small requested delays, use kthread instead of
 		 * timers and workqueue to achieve better latency.
 		 */
-		dc->worker = kthread_create(&flush_worker_fn, dc,
-					    "dm-delay-flush-worker");
+		dc->worker = kthread_run(&flush_worker_fn, dc, "dm-delay-flush-worker");
 		if (IS_ERR(dc->worker)) {
 			ret = PTR_ERR(dc->worker);
 			dc->worker = NULL;
@@ -309,14 +311,14 @@ static int delay_bio(struct delay_c *dc, struct delay_class *c, struct bio *bio)
 	delayed->context = dc;
 	delayed->expires = expires = jiffies + msecs_to_jiffies(c->delay);
 
-	mutex_lock(&delayed_bios_lock);
+	spin_lock(&dc->delayed_bios_lock);
 	if (unlikely(!dc->may_delay)) {
-		mutex_unlock(&delayed_bios_lock);
+		spin_unlock(&dc->delayed_bios_lock);
 		return DM_MAPIO_REMAPPED;
 	}
 	c->ops++;
 	list_add_tail(&delayed->list, &dc->delayed_bios);
-	mutex_unlock(&delayed_bios_lock);
+	spin_unlock(&dc->delayed_bios_lock);
 
 	if (delay_is_fast(dc))
 		wake_up_process(dc->worker);
@@ -330,12 +332,12 @@ static void delay_presuspend(struct dm_target *ti)
 {
 	struct delay_c *dc = ti->private;
 
-	mutex_lock(&delayed_bios_lock);
+	spin_lock(&dc->delayed_bios_lock);
 	dc->may_delay = false;
-	mutex_unlock(&delayed_bios_lock);
+	spin_unlock(&dc->delayed_bios_lock);
 
 	if (!delay_is_fast(dc))
-		del_timer_sync(&dc->delay_timer);
+		timer_delete(&dc->delay_timer);
 	flush_delayed_bios(dc, true);
 }
 
diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c
index 2c6fbd87363f..cc66a27c363a 100644
--- a/drivers/md/dm-table.c
+++ b/drivers/md/dm-table.c
@@ -1963,26 +1963,27 @@ int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
 	bool wc = false, fua = false;
 	int r;
 
-	/*
-	 * Copy table's limits to the DM device's request_queue
-	 */
-	q->limits = *limits;
-
 	if (dm_table_supports_nowait(t))
 		blk_queue_flag_set(QUEUE_FLAG_NOWAIT, q);
 	else
 		blk_queue_flag_clear(QUEUE_FLAG_NOWAIT, q);
 
 	if (!dm_table_supports_discards(t)) {
-		q->limits.max_discard_sectors = 0;
-		q->limits.max_hw_discard_sectors = 0;
-		q->limits.discard_granularity = 0;
-		q->limits.discard_alignment = 0;
-		q->limits.discard_misaligned = 0;
+		limits->max_hw_discard_sectors = 0;
+		limits->discard_granularity = 0;
+		limits->discard_alignment = 0;
+		limits->discard_misaligned = 0;
 	}
 
+	if (!dm_table_supports_write_zeroes(t))
+		limits->max_write_zeroes_sectors = 0;
+
 	if (!dm_table_supports_secure_erase(t))
-		q->limits.max_secure_erase_sectors = 0;
+		limits->max_secure_erase_sectors = 0;
+
+	r = queue_limits_set(q, limits);
+	if (r)
+		return r;
 
 	if (dm_table_supports_flush(t, (1UL << QUEUE_FLAG_WC))) {
 		wc = true;
@@ -2007,9 +2008,6 @@ int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
 	else
 		blk_queue_flag_set(QUEUE_FLAG_NONROT, q);
 
-	if (!dm_table_supports_write_zeroes(t))
-		q->limits.max_write_zeroes_sectors = 0;
-
 	dm_table_verify_integrity(t);
 
 	/*
@@ -2048,7 +2046,6 @@ int dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
 	}
 
 	dm_update_crypto_profile(q, t);
-	disk_update_readahead(t->md->disk);
 
 	/*
 	 * Check for request-based device is left to