[PATCH 1/4] rcu/kvfree: Switch to a generic linked list API

[PATCH 1/4] rcu/kvfree: Switch to a generic linked list API

[Uladzislau Rezki (Sony)]

To make a code more readable and less confusing switch
to a standard circular double linked list API. It allows
to simplify the code since basic list operations are well
defined and documented.

Please note, this patch does not introduce any functional
change it is only limited by refactoring of code.

Signed-off-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
---
 kernel/rcu/tree.c | 89 +++++++++++++++++++++++------------------------
 1 file changed, 43 insertions(+), 46 deletions(-)

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 76973d716921..74d6889dcc50 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -2900,13 +2900,13 @@ EXPORT_SYMBOL_GPL(call_rcu);
 
 /**
  * struct kvfree_rcu_bulk_data - single block to store kvfree_rcu() pointers
+ * @list: List node. All blocks are linked between each other
  * @nr_records: Number of active pointers in the array
- * @next: Next bulk object in the block chain
  * @records: Array of the kvfree_rcu() pointers
  */
 struct kvfree_rcu_bulk_data {
+	struct list_head list;
 	unsigned long nr_records;
-	struct kvfree_rcu_bulk_data *next;
 	void *records[];
 };
 
@@ -2922,21 +2922,21 @@ struct kvfree_rcu_bulk_data {
  * struct kfree_rcu_cpu_work - single batch of kfree_rcu() requests
  * @rcu_work: Let queue_rcu_work() invoke workqueue handler after grace period
  * @head_free: List of kfree_rcu() objects waiting for a grace period
- * @bkvhead_free: Bulk-List of kvfree_rcu() objects waiting for a grace period
+ * @bulk_head_free: Bulk-List of kvfree_rcu() objects waiting for a grace period
  * @krcp: Pointer to @kfree_rcu_cpu structure
  */
 
 struct kfree_rcu_cpu_work {
 	struct rcu_work rcu_work;
 	struct rcu_head *head_free;
-	struct kvfree_rcu_bulk_data *bkvhead_free[FREE_N_CHANNELS];
+	struct list_head bulk_head_free[FREE_N_CHANNELS];
 	struct kfree_rcu_cpu *krcp;
 };
 
 /**
  * struct kfree_rcu_cpu - batch up kfree_rcu() requests for RCU grace period
  * @head: List of kfree_rcu() objects not yet waiting for a grace period
- * @bkvhead: Bulk-List of kvfree_rcu() objects not yet waiting for a grace period
+ * @bulk_head: Bulk-List of kvfree_rcu() objects not yet waiting for a grace period
  * @krw_arr: Array of batches of kfree_rcu() objects waiting for a grace period
  * @lock: Synchronize access to this structure
  * @monitor_work: Promote @head to @head_free after KFREE_DRAIN_JIFFIES
@@ -2960,7 +2960,7 @@ struct kfree_rcu_cpu_work {
  */
 struct kfree_rcu_cpu {
 	struct rcu_head *head;
-	struct kvfree_rcu_bulk_data *bkvhead[FREE_N_CHANNELS];
+	struct list_head bulk_head[FREE_N_CHANNELS];
 	struct kfree_rcu_cpu_work krw_arr[KFREE_N_BATCHES];
 	raw_spinlock_t lock;
 	struct delayed_work monitor_work;
@@ -3055,12 +3055,13 @@ drain_page_cache(struct kfree_rcu_cpu *krcp)
 
 /*
  * This function is invoked in workqueue context after a grace period.
- * It frees all the objects queued on ->bkvhead_free or ->head_free.
+ * It frees all the objects queued on ->bulk_head_free or ->head_free.
  */
 static void kfree_rcu_work(struct work_struct *work)
 {
 	unsigned long flags;
-	struct kvfree_rcu_bulk_data *bkvhead[FREE_N_CHANNELS], *bnext;
+	struct kvfree_rcu_bulk_data *bnode, *n;
+	struct list_head bulk_head[FREE_N_CHANNELS];
 	struct rcu_head *head, *next;
 	struct kfree_rcu_cpu *krcp;
 	struct kfree_rcu_cpu_work *krwp;
@@ -3072,10 +3073,8 @@ static void kfree_rcu_work(struct work_struct *work)
 
 	raw_spin_lock_irqsave(&krcp->lock, flags);
 	// Channels 1 and 2.
-	for (i = 0; i < FREE_N_CHANNELS; i++) {
-		bkvhead[i] = krwp->bkvhead_free[i];
-		krwp->bkvhead_free[i] = NULL;
-	}
+	for (i = 0; i < FREE_N_CHANNELS; i++)
+		list_replace_init(&krwp->bulk_head_free[i], &bulk_head[i]);
 
 	// Channel 3.
 	head = krwp->head_free;
@@ -3084,36 +3083,33 @@ static void kfree_rcu_work(struct work_struct *work)
 
 	// Handle the first two channels.
 	for (i = 0; i < FREE_N_CHANNELS; i++) {
-		for (; bkvhead[i]; bkvhead[i] = bnext) {
-			bnext = bkvhead[i]->next;
-			debug_rcu_bhead_unqueue(bkvhead[i]);
+		list_for_each_entry_safe(bnode, n, &bulk_head[i], list) {
+			debug_rcu_bhead_unqueue(bnode);
 
 			rcu_lock_acquire(&rcu_callback_map);
 			if (i == 0) { // kmalloc() / kfree().
 				trace_rcu_invoke_kfree_bulk_callback(
-					rcu_state.name, bkvhead[i]->nr_records,
-					bkvhead[i]->records);
+					rcu_state.name, bnode->nr_records,
+					bnode->records);
 
-				kfree_bulk(bkvhead[i]->nr_records,
-					bkvhead[i]->records);
+				kfree_bulk(bnode->nr_records, bnode->records);
 			} else { // vmalloc() / vfree().
-				for (j = 0; j < bkvhead[i]->nr_records; j++) {
+				for (j = 0; j < bnode->nr_records; j++) {
 					trace_rcu_invoke_kvfree_callback(
-						rcu_state.name,
-						bkvhead[i]->records[j], 0);
+						rcu_state.name, bnode->records[j], 0);
 
-					vfree(bkvhead[i]->records[j]);
+					vfree(bnode->records[j]);
 				}
 			}
 			rcu_lock_release(&rcu_callback_map);
 
 			raw_spin_lock_irqsave(&krcp->lock, flags);
-			if (put_cached_bnode(krcp, bkvhead[i]))
-				bkvhead[i] = NULL;
+			if (put_cached_bnode(krcp, bnode))
+				bnode = NULL;
 			raw_spin_unlock_irqrestore(&krcp->lock, flags);
 
-			if (bkvhead[i])
-				free_page((unsigned long) bkvhead[i]);
+			if (bnode)
+				free_page((unsigned long) bnode);
 
 			cond_resched_tasks_rcu_qs();
 		}
@@ -3149,7 +3145,7 @@ need_offload_krc(struct kfree_rcu_cpu *krcp)
 	int i;
 
 	for (i = 0; i < FREE_N_CHANNELS; i++)
-		if (krcp->bkvhead[i])
+		if (!list_empty(&krcp->bulk_head[i]))
 			return true;
 
 	return !!krcp->head;
@@ -3186,21 +3182,20 @@ static void kfree_rcu_monitor(struct work_struct *work)
 	for (i = 0; i < KFREE_N_BATCHES; i++) {
 		struct kfree_rcu_cpu_work *krwp = &(krcp->krw_arr[i]);
 
-		// Try to detach bkvhead or head and attach it over any
+		// Try to detach bulk_head or head and attach it over any
 		// available corresponding free channel. It can be that
 		// a previous RCU batch is in progress, it means that
 		// immediately to queue another one is not possible so
 		// in that case the monitor work is rearmed.
-		if ((krcp->bkvhead[0] && !krwp->bkvhead_free[0]) ||
-			(krcp->bkvhead[1] && !krwp->bkvhead_free[1]) ||
+		if ((!list_empty(&krcp->bulk_head[0]) && list_empty(&krwp->bulk_head_free[0])) ||
+			(!list_empty(&krcp->bulk_head[1]) && list_empty(&krwp->bulk_head_free[1])) ||
 				(krcp->head && !krwp->head_free)) {
+
 			// Channel 1 corresponds to the SLAB-pointer bulk path.
 			// Channel 2 corresponds to vmalloc-pointer bulk path.
 			for (j = 0; j < FREE_N_CHANNELS; j++) {
-				if (!krwp->bkvhead_free[j]) {
-					krwp->bkvhead_free[j] = krcp->bkvhead[j];
-					krcp->bkvhead[j] = NULL;
-				}
+				if (list_empty(&krwp->bulk_head_free[j]))
+					list_replace_init(&krcp->bulk_head[j], &krwp->bulk_head_free[j]);
 			}
 
 			// Channel 3 corresponds to both SLAB and vmalloc
@@ -3312,10 +3307,11 @@ add_ptr_to_bulk_krc_lock(struct kfree_rcu_cpu **krcp,
 		return false;
 
 	idx = !!is_vmalloc_addr(ptr);
+	bnode = list_first_entry_or_null(&(*krcp)->bulk_head[idx],
+		struct kvfree_rcu_bulk_data, list);
 
 	/* Check if a new block is required. */
-	if (!(*krcp)->bkvhead[idx] ||
-			(*krcp)->bkvhead[idx]->nr_records == KVFREE_BULK_MAX_ENTR) {
+	if (!bnode || bnode->nr_records == KVFREE_BULK_MAX_ENTR) {
 		bnode = get_cached_bnode(*krcp);
 		if (!bnode && can_alloc) {
 			krc_this_cpu_unlock(*krcp, *flags);
@@ -3339,18 +3335,13 @@ add_ptr_to_bulk_krc_lock(struct kfree_rcu_cpu **krcp,
 		if (!bnode)
 			return false;
 
-		/* Initialize the new block. */
+		// Initialize the new block and attach it.
 		bnode->nr_records = 0;
-		bnode->next = (*krcp)->bkvhead[idx];
-
-		/* Attach it to the head. */
-		(*krcp)->bkvhead[idx] = bnode;
+		list_add(&bnode->list, &(*krcp)->bulk_head[idx]);
 	}
 
 	/* Finally insert. */
-	(*krcp)->bkvhead[idx]->records
-		[(*krcp)->bkvhead[idx]->nr_records++] = ptr;
-
+	bnode->records[bnode->nr_records++] = ptr;
 	return true;
 }
 
@@ -4790,7 +4781,7 @@ struct workqueue_struct *rcu_gp_wq;
 static void __init kfree_rcu_batch_init(void)
 {
 	int cpu;
-	int i;
+	int i, j;
 
 	/* Clamp it to [0:100] seconds interval. */
 	if (rcu_delay_page_cache_fill_msec < 0 ||
@@ -4810,8 +4801,14 @@ static void __init kfree_rcu_batch_init(void)
 		for (i = 0; i < KFREE_N_BATCHES; i++) {
 			INIT_RCU_WORK(&krcp->krw_arr[i].rcu_work, kfree_rcu_work);
 			krcp->krw_arr[i].krcp = krcp;
+
+			for (j = 0; j < FREE_N_CHANNELS; j++)
+				INIT_LIST_HEAD(&krcp->krw_arr[i].bulk_head_free[j]);
 		}
 
+		for (i = 0; i < FREE_N_CHANNELS; i++)
+			INIT_LIST_HEAD(&krcp->bulk_head[i]);
+
 		INIT_DELAYED_WORK(&krcp->monitor_work, kfree_rcu_monitor);
 		INIT_DELAYED_WORK(&krcp->page_cache_work, fill_page_cache_func);
 		krcp->initialized = true;
-- 
2.30.2

[PATCH 2/4] rcu/kvfree: Move bulk/list reclaim to separate functions

[Uladzislau Rezki (Sony)]

There are two different paths how a memory is reclaimed.
Currently it is open-coded what makes it a bit messy and
less easy to read.

Introduce two separate functions kvfree_rcu_list() and
kvfree_rcu_bulk() to cover two independent cases.

Please note, this patch does not introduce any functional
change it is only limited by refactoring of code.

Signed-off-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
---
 kernel/rcu/tree.c | 114 ++++++++++++++++++++++++++--------------------
 1 file changed, 65 insertions(+), 49 deletions(-)

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 74d6889dcc50..3b5f6036d884 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -3053,6 +3053,65 @@ drain_page_cache(struct kfree_rcu_cpu *krcp)
 	return freed;
 }
 
+static void
+kvfree_rcu_bulk(struct kfree_rcu_cpu *krcp,
+	struct kvfree_rcu_bulk_data *bnode, int idx)
+{
+	unsigned long flags;
+	int i;
+
+	debug_rcu_bhead_unqueue(bnode);
+
+	rcu_lock_acquire(&rcu_callback_map);
+	if (idx == 0) { // kmalloc() / kfree().
+		trace_rcu_invoke_kfree_bulk_callback(
+			rcu_state.name, bnode->nr_records,
+			bnode->records);
+
+		kfree_bulk(bnode->nr_records, bnode->records);
+	} else { // vmalloc() / vfree().
+		for (i = 0; i < bnode->nr_records; i++) {
+			trace_rcu_invoke_kvfree_callback(
+				rcu_state.name, bnode->records[i], 0);
+
+			vfree(bnode->records[i]);
+		}
+	}
+	rcu_lock_release(&rcu_callback_map);
+
+	raw_spin_lock_irqsave(&krcp->lock, flags);
+	if (put_cached_bnode(krcp, bnode))
+		bnode = NULL;
+	raw_spin_unlock_irqrestore(&krcp->lock, flags);
+
+	if (bnode)
+		free_page((unsigned long) bnode);
+
+	cond_resched_tasks_rcu_qs();
+}
+
+static void
+kvfree_rcu_list(struct rcu_head *head)
+{
+	struct rcu_head *next;
+
+	for (; head; head = next) {
+		unsigned long offset = (unsigned long)head->func;
+		void *ptr = (void *)head - offset;
+
+		next = head->next;
+		debug_rcu_head_unqueue((struct rcu_head *)ptr);
+		rcu_lock_acquire(&rcu_callback_map);
+		trace_rcu_invoke_kvfree_callback(rcu_state.name, head, offset);
+
+		if (!WARN_ON_ONCE(!__is_kvfree_rcu_offset(offset)))
+			kvfree(ptr);
+
+		rcu_lock_release(&rcu_callback_map);
+		cond_resched_tasks_rcu_qs();
+	}
+}
+
 /*
  * This function is invoked in workqueue context after a grace period.
  * It frees all the objects queued on ->bulk_head_free or ->head_free.
@@ -3062,10 +3121,10 @@ static void kfree_rcu_work(struct work_struct *work)
 	unsigned long flags;
 	struct kvfree_rcu_bulk_data *bnode, *n;
 	struct list_head bulk_head[FREE_N_CHANNELS];
-	struct rcu_head *head, *next;
+	struct rcu_head *head;
 	struct kfree_rcu_cpu *krcp;
 	struct kfree_rcu_cpu_work *krwp;
-	int i, j;
+	int i;
 
 	krwp = container_of(to_rcu_work(work),
 			    struct kfree_rcu_cpu_work, rcu_work);
@@ -3082,38 +3141,9 @@ static void kfree_rcu_work(struct work_struct *work)
 	raw_spin_unlock_irqrestore(&krcp->lock, flags);
 
 	// Handle the first two channels.
-	for (i = 0; i < FREE_N_CHANNELS; i++) {
-		list_for_each_entry_safe(bnode, n, &bulk_head[i], list) {
-			debug_rcu_bhead_unqueue(bnode);
-
-			rcu_lock_acquire(&rcu_callback_map);
-			if (i == 0) { // kmalloc() / kfree().
-				trace_rcu_invoke_kfree_bulk_callback(
-					rcu_state.name, bnode->nr_records,
-					bnode->records);
-
-				kfree_bulk(bnode->nr_records, bnode->records);
-			} else { // vmalloc() / vfree().
-				for (j = 0; j < bnode->nr_records; j++) {
-					trace_rcu_invoke_kvfree_callback(
-						rcu_state.name, bnode->records[j], 0);
-
-					vfree(bnode->records[j]);
-				}
-			}
-			rcu_lock_release(&rcu_callback_map);
-
-			raw_spin_lock_irqsave(&krcp->lock, flags);
-			if (put_cached_bnode(krcp, bnode))
-				bnode = NULL;
-			raw_spin_unlock_irqrestore(&krcp->lock, flags);
-
-			if (bnode)
-				free_page((unsigned long) bnode);
-
-			cond_resched_tasks_rcu_qs();
-		}
-	}
+	for (i = 0; i < FREE_N_CHANNELS; i++)
+		list_for_each_entry_safe(bnode, n, &bulk_head[i], list)
+			kvfree_rcu_bulk(krcp, bnode, i);
 
 	/*
 	 * This is used when the "bulk" path can not be used for the
@@ -3122,21 +3152,7 @@ static void kfree_rcu_work(struct work_struct *work)
 	 * queued on a linked list through their rcu_head structures.
 	 * This list is named "Channel 3".
 	 */
-	for (; head; head = next) {
-		unsigned long offset = (unsigned long)head->func;
-		void *ptr = (void *)head - offset;
-
-		next = head->next;
-		debug_rcu_head_unqueue((struct rcu_head *)ptr);
-		rcu_lock_acquire(&rcu_callback_map);
-		trace_rcu_invoke_kvfree_callback(rcu_state.name, head, offset);
-
-		if (!WARN_ON_ONCE(!__is_kvfree_rcu_offset(offset)))
-			kvfree(ptr);
-
-		rcu_lock_release(&rcu_callback_map);
-		cond_resched_tasks_rcu_qs();
-	}
+	kvfree_rcu_list(head);
 }
 
 static bool
-- 
2.30.2

[PATCH 3/4] rcu/kvfree: Move need_offload_krc() out of krcp->lock

[Uladzislau Rezki (Sony)]

Currently a need_offload_krc() function requires the krcp->lock
to be held because krcp->head can not be checked concurrently.

Fix it by updating the krcp->head using WRITE_ONCE() macro so
it becomes lock-free and safe for readers to see a valid data
without any locking.

Signed-off-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
---
 kernel/rcu/tree.c | 11 ++++-------
 1 file changed, 4 insertions(+), 7 deletions(-)

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 3b5f6036d884..f68ddbef2a33 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -3218,7 +3218,7 @@ static void kfree_rcu_monitor(struct work_struct *work)
 			// objects queued on the linked list.
 			if (!krwp->head_free) {
 				krwp->head_free = krcp->head;
-				krcp->head = NULL;
+				WRITE_ONCE(krcp->head, NULL);
 			}
 
 			WRITE_ONCE(krcp->count, 0);
@@ -3232,6 +3232,8 @@ static void kfree_rcu_monitor(struct work_struct *work)
 		}
 	}
 
+	raw_spin_unlock_irqrestore(&krcp->lock, flags);
+
 	// If there is nothing to detach, it means that our job is
 	// successfully done here. In case of having at least one
 	// of the channels that is still busy we should rearm the
@@ -3239,8 +3241,6 @@ static void kfree_rcu_monitor(struct work_struct *work)
 	// still in progress.
 	if (need_offload_krc(krcp))
 		schedule_delayed_monitor_work(krcp);
-
-	raw_spin_unlock_irqrestore(&krcp->lock, flags);
 }
 
 static enum hrtimer_restart
@@ -3415,7 +3415,7 @@ void kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func)
 
 		head->func = func;
 		head->next = krcp->head;
-		krcp->head = head;
+		WRITE_ONCE(krcp->head, head);
 		success = true;
 	}
 
@@ -3492,15 +3492,12 @@ static struct shrinker kfree_rcu_shrinker = {
 void __init kfree_rcu_scheduler_running(void)
 {
 	int cpu;
-	unsigned long flags;
 
 	for_each_possible_cpu(cpu) {
 		struct kfree_rcu_cpu *krcp = per_cpu_ptr(&krc, cpu);
 
-		raw_spin_lock_irqsave(&krcp->lock, flags);
 		if (need_offload_krc(krcp))
 			schedule_delayed_monitor_work(krcp);
-		raw_spin_unlock_irqrestore(&krcp->lock, flags);
 	}
 }
 
-- 
2.30.2

[PATCH 4/4] rcu/kvfree: Use a polled API to speedup a reclaim process

[Uladzislau Rezki (Sony)]

Currently all objects placed into a batch require a full GP
after passing which objects in a batch are eligible to be
freed.

The problem is that many pointers may already passed several
GP sequences so there is no need for them in extra delay and
such objects can be reclaimed right away without waiting.

In order to reduce a memory footprint this patch introduces
a per-page-grace-period-controlling mechanism. It allows us
to distinguish pointers for which a grace period is passed
and for which not.

A reclaim thread in its turn frees a memory in a reverse
order starting from a tail because a GP is likely passed
for objects in a page. If a page with a GP sequence in a
list hits a condition when a GP is not ready we bail out
requesting one more grace period in order to complete a
drain process for left pages.

Test example:

kvm.sh --memory 10G --torture rcuscale --allcpus --duration 1 \
  --kconfig CONFIG_NR_CPUS=64 \
  --kconfig CONFIG_RCU_NOCB_CPU=y \
  --kconfig CONFIG_RCU_NOCB_CPU_DEFAULT_ALL=y \
  --kconfig CONFIG_RCU_LAZY=n \
  --bootargs "rcuscale.kfree_rcu_test=1 rcuscale.kfree_nthreads=16 \
  rcuscale.holdoff=20 rcuscale.kfree_loops=10000 \
  torture.disable_onoff_at_boot" --trust-make

Total time taken by all kfree'ers: 8535693700 ns, loops: 10000, batches: 1188, memory footprint: 2248MB
Total time taken by all kfree'ers: 8466933582 ns, loops: 10000, batches: 1157, memory footprint: 2820MB
Total time taken by all kfree'ers: 5375602446 ns, loops: 10000, batches: 1130, memory footprint: 6502MB
Total time taken by all kfree'ers: 7523283832 ns, loops: 10000, batches: 1006, memory footprint: 3343MB
Total time taken by all kfree'ers: 6459171956 ns, loops: 10000, batches: 1150, memory footprint: 6549MB

Total time taken by all kfree'ers: 8560060176 ns, loops: 10000, batches: 1787, memory footprint: 61MB
Total time taken by all kfree'ers: 8573885501 ns, loops: 10000, batches: 1777, memory footprint: 93MB
Total time taken by all kfree'ers: 8320000202 ns, loops: 10000, batches: 1727, memory footprint: 66MB
Total time taken by all kfree'ers: 8552718794 ns, loops: 10000, batches: 1790, memory footprint: 75MB
Total time taken by all kfree'ers: 8601368792 ns, loops: 10000, batches: 1724, memory footprint: 62MB

Signed-off-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
---
 kernel/rcu/tree.c | 47 +++++++++++++++++++++++++++++++++++++++--------
 1 file changed, 39 insertions(+), 8 deletions(-)

diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index f68ddbef2a33..b41241994672 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -2901,11 +2901,13 @@ EXPORT_SYMBOL_GPL(call_rcu);
 /**
  * struct kvfree_rcu_bulk_data - single block to store kvfree_rcu() pointers
  * @list: List node. All blocks are linked between each other
+ * @gp_snap: Snapshot of RCU state for objects placed to this bulk
  * @nr_records: Number of active pointers in the array
  * @records: Array of the kvfree_rcu() pointers
  */
 struct kvfree_rcu_bulk_data {
 	struct list_head list;
+	unsigned long gp_snap;
 	unsigned long nr_records;
 	void *records[];
 };
@@ -2922,13 +2924,15 @@ struct kvfree_rcu_bulk_data {
  * struct kfree_rcu_cpu_work - single batch of kfree_rcu() requests
  * @rcu_work: Let queue_rcu_work() invoke workqueue handler after grace period
  * @head_free: List of kfree_rcu() objects waiting for a grace period
+ * @head_free_gp_snap: Snapshot of RCU state for objects placed to "@head_free"
  * @bulk_head_free: Bulk-List of kvfree_rcu() objects waiting for a grace period
  * @krcp: Pointer to @kfree_rcu_cpu structure
  */
 
 struct kfree_rcu_cpu_work {
-	struct rcu_work rcu_work;
+	struct work_struct rcu_work;
 	struct rcu_head *head_free;
+	unsigned long head_free_gp_snap;
 	struct list_head bulk_head_free[FREE_N_CHANNELS];
 	struct kfree_rcu_cpu *krcp;
 };
@@ -3124,10 +3128,11 @@ static void kfree_rcu_work(struct work_struct *work)
 	struct rcu_head *head;
 	struct kfree_rcu_cpu *krcp;
 	struct kfree_rcu_cpu_work *krwp;
+	unsigned long head_free_gp_snap;
 	int i;
 
-	krwp = container_of(to_rcu_work(work),
-			    struct kfree_rcu_cpu_work, rcu_work);
+	krwp = container_of(work,
+		struct kfree_rcu_cpu_work, rcu_work);
 	krcp = krwp->krcp;
 
 	raw_spin_lock_irqsave(&krcp->lock, flags);
@@ -3138,12 +3143,29 @@ static void kfree_rcu_work(struct work_struct *work)
 	// Channel 3.
 	head = krwp->head_free;
 	krwp->head_free = NULL;
+	head_free_gp_snap = krwp->head_free_gp_snap;
 	raw_spin_unlock_irqrestore(&krcp->lock, flags);
 
 	// Handle the first two channels.
-	for (i = 0; i < FREE_N_CHANNELS; i++)
+	for (i = 0; i < FREE_N_CHANNELS; i++) {
+		// Start from the tail page, so a GP is likely passed for it.
+		list_for_each_entry_safe_reverse(bnode, n, &bulk_head[i], list) {
+			// Not yet ready? Bail out since we need one more GP.
+			if (!poll_state_synchronize_rcu(bnode->gp_snap))
+				break;
+
+			list_del_init(&bnode->list);
+			kvfree_rcu_bulk(krcp, bnode, i);
+		}
+
+		// Please note a request for one more extra GP can
+		// occur only once for all objects in this batch.
+		if (!list_empty(&bulk_head[i]))
+			synchronize_rcu();
+
 		list_for_each_entry_safe(bnode, n, &bulk_head[i], list)
 			kvfree_rcu_bulk(krcp, bnode, i);
+	}
 
 	/*
 	 * This is used when the "bulk" path can not be used for the
@@ -3152,7 +3174,10 @@ static void kfree_rcu_work(struct work_struct *work)
 	 * queued on a linked list through their rcu_head structures.
 	 * This list is named "Channel 3".
 	 */
-	kvfree_rcu_list(head);
+	if (head) {
+		cond_synchronize_rcu(head_free_gp_snap);
+		kvfree_rcu_list(head);
+	}
 }
 
 static bool
@@ -3219,6 +3244,11 @@ static void kfree_rcu_monitor(struct work_struct *work)
 			if (!krwp->head_free) {
 				krwp->head_free = krcp->head;
 				WRITE_ONCE(krcp->head, NULL);
+
+				// Take a snapshot for this krwp. Please note no more
+				// any objects can be added to attached head_free channel
+				// therefore fixate a GP for it here.
+				krwp->head_free_gp_snap = get_state_synchronize_rcu();
 			}
 
 			WRITE_ONCE(krcp->count, 0);
@@ -3228,7 +3258,7 @@ static void kfree_rcu_monitor(struct work_struct *work)
 			// be that the work is in the pending state when
 			// channels have been detached following by each
 			// other.
-			queue_rcu_work(system_wq, &krwp->rcu_work);
+			queue_work(system_wq, &krwp->rcu_work);
 		}
 	}
 
@@ -3356,8 +3386,9 @@ add_ptr_to_bulk_krc_lock(struct kfree_rcu_cpu **krcp,
 		list_add(&bnode->list, &(*krcp)->bulk_head[idx]);
 	}
 
-	/* Finally insert. */
+	// Finally insert and update the GP for this page.
 	bnode->records[bnode->nr_records++] = ptr;
+	bnode->gp_snap = get_state_synchronize_rcu();
 	return true;
 }
 
@@ -4812,7 +4843,7 @@ static void __init kfree_rcu_batch_init(void)
 		struct kfree_rcu_cpu *krcp = per_cpu_ptr(&krc, cpu);
 
 		for (i = 0; i < KFREE_N_BATCHES; i++) {
-			INIT_RCU_WORK(&krcp->krw_arr[i].rcu_work, kfree_rcu_work);
+			INIT_WORK(&krcp->krw_arr[i].rcu_work, kfree_rcu_work);
 			krcp->krw_arr[i].krcp = krcp;
 
 			for (j = 0; j < FREE_N_CHANNELS; j++)
-- 
2.30.2