Date: Tue, 16 Jun 2015 11:41:51 -0400
From: Steven Rostedt <rostedt@goodmis.org>
To: Daniel Wagner <wagi@monom.org>
Cc: paulmck@linux.vnet.ibm.com, Alexei Starovoitov <ast@plumgrid.com>,
        Daniel Wagner <daniel.wagner@bmw-carit.de>,
        LKML <linux-kernel@vger.kernel.org>
Subject: Re: call_rcu from trace_preempt
Message-ID: <20150616114151.3681a9e8@gandalf.local.home>
In-Reply-To: <558018DD.1080701@monom.org>
References: <557F509D.2000509@plumgrid.com>
	<20150615230702.GB3913@linux.vnet.ibm.com>
	<557F7764.5060707@plumgrid.com>
	<20150616021458.GE3913@linux.vnet.ibm.com>
	<557FB7E1.6080004@plumgrid.com>
	<20150616122733.GG3913@linux.vnet.ibm.com>
	<558018DD.1080701@monom.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 2921
Lines: 110

On Tue, 16 Jun 2015 14:38:53 +0200
Daniel Wagner <wagi@monom.org> wrote:

> diff --git a/kernel/bpf/hashtab.c b/kernel/bpf/hashtab.c
> index 83c209d..8d73be3 100644
> --- a/kernel/bpf/hashtab.c
> +++ b/kernel/bpf/hashtab.c
> @@ -13,6 +13,8 @@
>  #include <linux/jhash.h>
>  #include <linux/filter.h>
>  #include <linux/vmalloc.h>
> +#include <linux/kthread.h>
> +#include <linux/spinlock.h>
> 
>  struct bpf_htab {
>  	struct bpf_map map;
> @@ -27,10 +29,14 @@ struct bpf_htab {
>  struct htab_elem {
>  	struct hlist_node hash_node;
>  	struct rcu_head rcu;
> +	struct list_head list;
>  	u32 hash;
>  	char key[0] __aligned(8);
>  };
> 
> +static LIST_HEAD(elem_freelist);
> +static DEFINE_SPINLOCK(elem_freelist_lock);
> +
>  /* Called from syscall */
>  static struct bpf_map *htab_map_alloc(union bpf_attr *attr)
>  {
> @@ -228,6 +234,7 @@ static int htab_map_update_elem(struct bpf_map *map,
> void *key, void *value,
>  	memcpy(l_new->key + round_up(key_size, 8), value, map->value_size);
> 
>  	l_new->hash = htab_map_hash(l_new->key, key_size);
> +	INIT_LIST_HEAD(&l_new->list);
> 
>  	/* bpf_map_update_elem() can be called in_irq() */
>  	spin_lock_irqsave(&htab->lock, flags);
> @@ -300,11 +307,17 @@ static int htab_map_delete_elem(struct bpf_map
> *map, void *key)
>  	if (l) {
>  		hlist_del_rcu(&l->hash_node);
>  		htab->count--;
> -		kfree_rcu(l, rcu);
> +		/* kfree_rcu(l, rcu); */

So this kfree_rcu() is only being used to defer a free, and has nothing
to do with having to free 'l' from rcu?

>  		ret = 0;
>  	}
> 
>  	spin_unlock_irqrestore(&htab->lock, flags);
> +
> +	if (l) {
> +		spin_lock_irqsave(&elem_freelist_lock, flags);
> +		list_add(&l->list, &elem_freelist);
> +		spin_unlock_irqrestore(&elem_freelist_lock, flags);
> +	}
>  	return ret;
>  }
> 
> @@ -359,9 +372,31 @@ static struct bpf_map_type_list htab_type
> __read_mostly = {
>  	.type = BPF_MAP_TYPE_HASH,
>  };
> 
> +static int free_thread(void *arg)
> +{
> +	unsigned long flags;
> +	struct htab_elem *l;
> +
> +	while (!kthread_should_stop()) {
> +		spin_lock_irqsave(&elem_freelist_lock, flags);
> +		while (!list_empty(&elem_freelist)) {
> +			l = list_entry(elem_freelist.next,
> +				struct htab_elem, list);
> +			list_del(&l->list);
> +			kfree(l);
> +		}
> +		spin_unlock_irqrestore(&elem_freelist_lock, flags);

Wow! This is burning up CPU isn't it?

If you just need to delay the kfree, why not use irq_work for that job?

-- Steve

> +	}
> +
> +	return 0;
> +}
> +
>  static int __init register_htab_map(void)
>  {
>  	bpf_register_map_type(&htab_type);
> +
> +	kthread_run(free_thread, NULL, "free_thread");
> +
>  	return 0;
>  }
>  late_initcall(register_htab_map);

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/