Skip to content

Commit

Permalink
rcu/kvfree: Add kvfree_rcu_barrier() API
Browse files Browse the repository at this point in the history
Add a kvfree_rcu_barrier() function. It waits until all
in-flight pointers are freed over RCU machinery. It does
not wait any GP completion and it is within its right to
return immediately if there are no outstanding pointers.

This function is useful when there is a need to guarantee
that a memory is fully freed before destroying memory caches.
For example, during unloading a kernel module.

Signed-off-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
  • Loading branch information
urezki authored and tehcaster committed Aug 27, 2024
1 parent 2eb14c1 commit 2b55d6a
Show file tree
Hide file tree
Showing 3 changed files with 107 additions and 8 deletions.
5 changes: 5 additions & 0 deletions include/linux/rcutiny.h
Original file line number Diff line number Diff line change
Expand Up @@ -111,6 +111,11 @@ static inline void __kvfree_call_rcu(struct rcu_head *head, void *ptr)
kvfree(ptr);
}

static inline void kvfree_rcu_barrier(void)
{
rcu_barrier();
}

#ifdef CONFIG_KASAN_GENERIC
void kvfree_call_rcu(struct rcu_head *head, void *ptr);
#else
Expand Down
1 change: 1 addition & 0 deletions include/linux/rcutree.h
Original file line number Diff line number Diff line change
Expand Up @@ -35,6 +35,7 @@ static inline void rcu_virt_note_context_switch(void)

void synchronize_rcu_expedited(void);
void kvfree_call_rcu(struct rcu_head *head, void *ptr);
void kvfree_rcu_barrier(void);

void rcu_barrier(void);
void rcu_momentary_dyntick_idle(void);
Expand Down
109 changes: 101 additions & 8 deletions kernel/rcu/tree.c
Original file line number Diff line number Diff line change
Expand Up @@ -3584,18 +3584,15 @@ kvfree_rcu_drain_ready(struct kfree_rcu_cpu *krcp)
}

/*
* This function is invoked after the KFREE_DRAIN_JIFFIES timeout.
* Return: %true if a work is queued, %false otherwise.
*/
static void kfree_rcu_monitor(struct work_struct *work)
static bool
kvfree_rcu_queue_batch(struct kfree_rcu_cpu *krcp)
{
struct kfree_rcu_cpu *krcp = container_of(work,
struct kfree_rcu_cpu, monitor_work.work);
unsigned long flags;
bool queued = false;
int i, j;

// Drain ready for reclaim.
kvfree_rcu_drain_ready(krcp);

raw_spin_lock_irqsave(&krcp->lock, flags);

// Attempt to start a new batch.
Expand Down Expand Up @@ -3634,11 +3631,27 @@ 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);
queued = queue_rcu_work(system_wq, &krwp->rcu_work);
}
}

raw_spin_unlock_irqrestore(&krcp->lock, flags);
return queued;
}

/*
* This function is invoked after the KFREE_DRAIN_JIFFIES timeout.
*/
static void kfree_rcu_monitor(struct work_struct *work)
{
struct kfree_rcu_cpu *krcp = container_of(work,
struct kfree_rcu_cpu, monitor_work.work);

// Drain ready for reclaim.
kvfree_rcu_drain_ready(krcp);

// Queue a batch for a rest.
kvfree_rcu_queue_batch(krcp);

// If there is nothing to detach, it means that our job is
// successfully done here. In case of having at least one
Expand Down Expand Up @@ -3859,6 +3872,86 @@ void kvfree_call_rcu(struct rcu_head *head, void *ptr)
}
EXPORT_SYMBOL_GPL(kvfree_call_rcu);

/**
* kvfree_rcu_barrier - Wait until all in-flight kvfree_rcu() complete.
*
* Note that a single argument of kvfree_rcu() call has a slow path that
* triggers synchronize_rcu() following by freeing a pointer. It is done
* before the return from the function. Therefore for any single-argument
* call that will result in a kfree() to a cache that is to be destroyed
* during module exit, it is developer's responsibility to ensure that all
* such calls have returned before the call to kmem_cache_destroy().
*/
void kvfree_rcu_barrier(void)
{
struct kfree_rcu_cpu_work *krwp;
struct kfree_rcu_cpu *krcp;
bool queued;
int i, cpu;

/*
* Firstly we detach objects and queue them over an RCU-batch
* for all CPUs. Finally queued works are flushed for each CPU.
*
* Please note. If there are outstanding batches for a particular
* CPU, those have to be finished first following by queuing a new.
*/
for_each_possible_cpu(cpu) {
krcp = per_cpu_ptr(&krc, cpu);

/*
* Check if this CPU has any objects which have been queued for a
* new GP completion. If not(means nothing to detach), we are done
* with it. If any batch is pending/running for this "krcp", below
* per-cpu flush_rcu_work() waits its completion(see last step).
*/
if (!need_offload_krc(krcp))
continue;

while (1) {
/*
* If we are not able to queue a new RCU work it means:
* - batches for this CPU are still in flight which should
* be flushed first and then repeat;
* - no objects to detach, because of concurrency.
*/
queued = kvfree_rcu_queue_batch(krcp);

/*
* Bail out, if there is no need to offload this "krcp"
* anymore. As noted earlier it can run concurrently.
*/
if (queued || !need_offload_krc(krcp))
break;

/* There are ongoing batches. */
for (i = 0; i < KFREE_N_BATCHES; i++) {
krwp = &(krcp->krw_arr[i]);
flush_rcu_work(&krwp->rcu_work);
}
}
}

/*
* Now we guarantee that all objects are flushed.
*/
for_each_possible_cpu(cpu) {
krcp = per_cpu_ptr(&krc, cpu);

/*
* A monitor work can drain ready to reclaim objects
* directly. Wait its completion if running or pending.
*/
cancel_delayed_work_sync(&krcp->monitor_work);

for (i = 0; i < KFREE_N_BATCHES; i++) {
krwp = &(krcp->krw_arr[i]);
flush_rcu_work(&krwp->rcu_work);
}
}
}
EXPORT_SYMBOL_GPL(kvfree_rcu_barrier);

static unsigned long
kfree_rcu_shrink_count(struct shrinker *shrink, struct shrink_control *sc)
{
Expand Down

0 comments on commit 2b55d6a

Please sign in to comment.