[v2] ocfs2/cluster: unlock the o2hb_live_lock before the o2nm_depend_item()
diff mbox

Message ID 59FAB46C.9050703@huawei.com
State New
Headers show

Commit Message

zhendong chen Nov. 2, 2017, 6 a.m. UTC
In the following situation, the down_write() will be called under
the spin_lock(), which may lead a soft lockup:
o2hb_region_inc_user
 spin_lock(&o2hb_live_lock)
  o2hb_region_pin
   o2nm_depend_item
    configfs_depend_item
     inode_lock
      down_write
      -->here may sleep and reschedule

So we should unlock the o2hb_live_lock before the o2nm_depend_item(), and
get item reference in advance to prevent the region to be released.

Signed-off-by: Alex Chen <alex.chen@huawei.com>
Reviewed-by: Yiwen Jiang <jiangyiwen@huawei.com>
Reviewed-by: Jun Piao <piaojun@huawei.com>
---
 fs/ocfs2/cluster/heartbeat.c | 7 +++++++
 1 file changed, 7 insertions(+)

Comments

Joseph Qi Nov. 6, 2017, 9:37 a.m. UTC | #1
Hi Alex,

For local heartbeat mode, I understand that if the region is not found,
o2hb_live_lock won't be released. And if the region is found, it doesn't
have to iterate the next region. So I agree with your fix in this case.

But I still don't get how to make sure the safe iteration in case of
global heartbeat mode.

Thanks,
Joseph

On 17/11/2 14:00, alex chen wrote:
> In the following situation, the down_write() will be called under
> the spin_lock(), which may lead a soft lockup:
> o2hb_region_inc_user
>  spin_lock(&o2hb_live_lock)
>   o2hb_region_pin
>    o2nm_depend_item
>     configfs_depend_item
>      inode_lock
>       down_write
>       -->here may sleep and reschedule
> 
> So we should unlock the o2hb_live_lock before the o2nm_depend_item(), and
> get item reference in advance to prevent the region to be released.
> 
> Signed-off-by: Alex Chen <alex.chen@huawei.com>
> Reviewed-by: Yiwen Jiang <jiangyiwen@huawei.com>
> Reviewed-by: Jun Piao <piaojun@huawei.com>
> ---
>  fs/ocfs2/cluster/heartbeat.c | 7 +++++++
>  1 file changed, 7 insertions(+)
> 
> diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c
> index d020604..07b2fdc 100644
> --- a/fs/ocfs2/cluster/heartbeat.c
> +++ b/fs/ocfs2/cluster/heartbeat.c
> @@ -2399,8 +2399,15 @@ static int o2hb_region_pin(const char *region_uuid)
>  		if (reg->hr_item_pinned || reg->hr_item_dropped)
>  			goto skip_pin;
> 
> +		config_item_get(&reg->hr_item);
> +		spin_unlock(&o2hb_live_lock);
> +
>  		/* Ignore ENOENT only for local hb (userdlm domain) */
>  		ret = o2nm_depend_item(&reg->hr_item);
> +
> +		config_item_put(&reg->hr_item);
> +		spin_lock(&o2hb_live_lock);
> +
>  		if (!ret) {
>  			mlog(ML_CLUSTER, "Pin region %s\n", uuid);
>  			reg->hr_item_pinned = 1;
>
zhendong chen Nov. 7, 2017, 2:30 a.m. UTC | #2
Hi Joseph,

Thanks for your reply.

For the global heartbeat mode, the concurrently access to reg->hr_item_pinned
may be happened after unlocking. I will modify this patch and send the patch v3.

Thanks,
Alex

On 2017/11/6 17:37, Joseph Qi wrote:
> Hi Alex,
> 
> For local heartbeat mode, I understand that if the region is not found,
> o2hb_live_lock won't be released. And if the region is found, it doesn't
> have to iterate the next region. So I agree with your fix in this case.
> 
> But I still don't get how to make sure the safe iteration in case of
> global heartbeat mode.
> 
> Thanks,
> Joseph
> 
> On 17/11/2 14:00, alex chen wrote:
>> In the following situation, the down_write() will be called under
>> the spin_lock(), which may lead a soft lockup:
>> o2hb_region_inc_user
>>  spin_lock(&o2hb_live_lock)
>>   o2hb_region_pin
>>    o2nm_depend_item
>>     configfs_depend_item
>>      inode_lock
>>       down_write
>>       -->here may sleep and reschedule
>>
>> So we should unlock the o2hb_live_lock before the o2nm_depend_item(), and
>> get item reference in advance to prevent the region to be released.
>>
>> Signed-off-by: Alex Chen <alex.chen@huawei.com>
>> Reviewed-by: Yiwen Jiang <jiangyiwen@huawei.com>
>> Reviewed-by: Jun Piao <piaojun@huawei.com>
>> ---
>>  fs/ocfs2/cluster/heartbeat.c | 7 +++++++
>>  1 file changed, 7 insertions(+)
>>
>> diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c
>> index d020604..07b2fdc 100644
>> --- a/fs/ocfs2/cluster/heartbeat.c
>> +++ b/fs/ocfs2/cluster/heartbeat.c
>> @@ -2399,8 +2399,15 @@ static int o2hb_region_pin(const char *region_uuid)
>>  		if (reg->hr_item_pinned || reg->hr_item_dropped)
>>  			goto skip_pin;
>>
>> +		config_item_get(&reg->hr_item);
>> +		spin_unlock(&o2hb_live_lock);
>> +
>>  		/* Ignore ENOENT only for local hb (userdlm domain) */
>>  		ret = o2nm_depend_item(&reg->hr_item);
>> +
>> +		config_item_put(&reg->hr_item);
>> +		spin_lock(&o2hb_live_lock);
>> +
>>  		if (!ret) {
>>  			mlog(ML_CLUSTER, "Pin region %s\n", uuid);
>>  			reg->hr_item_pinned = 1;
>>

Patch
diff mbox

diff --git a/fs/ocfs2/cluster/heartbeat.c b/fs/ocfs2/cluster/heartbeat.c
index d020604..07b2fdc 100644
--- a/fs/ocfs2/cluster/heartbeat.c
+++ b/fs/ocfs2/cluster/heartbeat.c
@@ -2399,8 +2399,15 @@  static int o2hb_region_pin(const char *region_uuid)
 		if (reg->hr_item_pinned || reg->hr_item_dropped)
 			goto skip_pin;

+		config_item_get(&reg->hr_item);
+		spin_unlock(&o2hb_live_lock);
+
 		/* Ignore ENOENT only for local hb (userdlm domain) */
 		ret = o2nm_depend_item(&reg->hr_item);
+
+		config_item_put(&reg->hr_item);
+		spin_lock(&o2hb_live_lock);
+
 		if (!ret) {
 			mlog(ML_CLUSTER, "Pin region %s\n", uuid);
 			reg->hr_item_pinned = 1;