[Ocfs2-devel] [PATCH] ocfs2: do not set OCFS2_LOCK_UPCONVERT_FINISHING if nonblocking lock can not be granted at once

[Junxiao Bi <junxiao.bi@oracle.com>]

On 11/11/2014 09:41 AM, Xue jiufei wrote:
> Hi JunXiao,
> On 2014/11/10 9:52, Junxiao Bi wrote:
>> On 11/07/2014 05:27 PM, Xue jiufei wrote:
>>> On 2014/11/7 15:03, Junxiao Bi wrote:
>>>> On 11/07/2014 02:36 PM, Xue jiufei wrote:
>>>>> Hi Junxiao
>>>>>
>>>>> On 2014/11/7 13:54, Junxiao Bi wrote:
>>>>>> Hi Jiufei,
>>>>>>
>>>>>> On 11/07/2014 11:41 AM, Xue jiufei wrote:
>>>>>>> Function ocfs2_readpages() use nonblocking flag to avoid page lock
>>>>>>> inversion. It will trigger cluster hang because that flag
>>>>>>> OCFS2_LOCK_UPCONVERT_FINISHING is not cleared if nonblocking lock
>>>>>>> cannot be granted at once. The flag would prevent dc thread from
>>>>>>> downconverting. So other nodes cannot acheive this lockres for ever.
>>>>>>>
>>>>>>> So we should not set OCFS2_LOCK_UPCONVERT_FINISHING when receiving ast
>>>>>>> if nonblocking lock had already returned.
>>>>>>>
>>>>>>
>>>>>> Use OCFS2_LOCK_NONBLOCK_FINISHED can't fix this issue. Like there are
>>>>>> two processes doing non-blocking lock in the following order.
>>>>> I am wondering why these two processes can lock in such order.
>>>>
>>>> Yes, you are right. LOCK_BUSY flag stop this race. Your fix works. But
>>>> just as i mentioned, i think for non-blocking lock, no need do cluster
>>>> wild locking, because AST is async, before it reached, most of the time,
>>>> the non-blocking lock had returned fail.
>>> We have added some debug information in function dlm_proxy_ast_handler()
>>> and dlm_send_remote_lock_request(), and run race testcase before.
>>> The result shows that the probability that AST message came earlier
>>> than returned status of locking is about 1/4. 
>> That may depend on how your race testcase is made.
>>>
>>> In addition, if non-blocking lock do not do cluster wide lock, ocfs2_readpages()
>>> will never succeed, 
>> Jiufei, can you explain why?
>>
> If node A takes EX lock, and node B call ocfs2_readpages(), which do not do cluster
> wide lock, master will not notice node A to downconvert. So even node B call ocfs2_readpages()
> many times, it will not succeed.
Not sure whether this will happen. As i thought, if there is read ahead
ops for this file, there should also be read op which can trigger the
down convert. Any way i think we'd better not depend on this.

Reviewed-by: Junxiao Bi <junxiao.bi@oracle.com>

> 
>> Thanks,
>> Junxiao.
>>
>> which is not we wanted.
>>>
>>> Thanks,
>>> Jiufei
>>>>
>>>> Thanks,
>>>> Junxiao.
>>>>>> 1. Process A issue non-blocking lock fail, set OCFS2_LOCK_NONBLOCK_FINISHED.
>>>>> We had set OCFS2_LOCK_BUSY flag before process A call ocfs2_dlm_lock(),
>>>>>> 2. Process B issue non-blocking lock fail, set
>>>>>> OCFS2_LOCK_NONBLOCK_FINISHED again.
>>>>> So process B must wait until ast for Process A returns and clear OCFS2_LOCK_BUSY.
>>>>>
>>>>>> 3. Ast for process A come, clear OCFS2_LOCK_NONBLOCK_FINISHED.
>>>>>> 4. Ast for process B come, set OCFS2_LOCK_UPCONVERT_FINISHING.
>>>>>>
>>>>>> So down convert can still be blocked.
>>>>>>
>>>>>> To fix this issue, I think no need to call ocfs2_dlm_lock() for
>>>>>> non-blocking lock, just check local lockres->l_level is enough, if
>>>>>> request level equal or smaller than it, lock success, or lock fail.
>>>>>>
>>>>>> Thanks,
>>>>>> Junxiao.
>>>>>>
>>>>> Thanks,
>>>>> jiufei.
>>>>>
>>>>>>> Signed-off-by: joyce.xue <xuejiufei@huawei.com>
>>>>>>> ---
>>>>>>>  fs/ocfs2/dlmglue.c | 37 +++++++++++++++++++++++++++++++------
>>>>>>>  fs/ocfs2/ocfs2.h   |  6 ++++++
>>>>>>>  2 files changed, 37 insertions(+), 6 deletions(-)
>>>>>>>
>>>>>>> diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c
>>>>>>> index 21262f2..1ddb6f1 100644
>>>>>>> --- a/fs/ocfs2/dlmglue.c
>>>>>>> +++ b/fs/ocfs2/dlmglue.c
>>>>>>> @@ -861,8 +861,13 @@ static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lo
>>>>>>>  	 * We set the OCFS2_LOCK_UPCONVERT_FINISHING flag before clearing
>>>>>>>  	 * the OCFS2_LOCK_BUSY flag to prevent the dc thread from
>>>>>>>  	 * downconverting the lock before the upconvert has fully completed.
>>>>>>> +	 * Do not prevent the dc thread from downconverting if NONBLOCK lock
>>>>>>> +	 * had already returned.
>>>>>>>  	 */
>>>>>>> -	lockres_or_flags(lockres, OCFS2_LOCK_UPCONVERT_FINISHING);
>>>>>>> +	if (!(lockres->l_flags & OCFS2_LOCK_NONBLOCK_FINISHED))
>>>>>>> +		lockres_or_flags(lockres, OCFS2_LOCK_UPCONVERT_FINISHING);
>>>>>>> +	else
>>>>>>> +		lockres_clear_flags(lockres, OCFS2_LOCK_NONBLOCK_FINISHED);
>>>>>>>  
>>>>>>>  	lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
>>>>>>>  }
>>>>>>> @@ -1324,13 +1329,12 @@ static void lockres_add_mask_waiter(struct ocfs2_lock_res *lockres,
>>>>>>>  
>>>>>>>  /* returns 0 if the mw that was removed was already satisfied, -EBUSY
>>>>>>>   * if the mask still hadn't reached its goal */
>>>>>>> -static int lockres_remove_mask_waiter(struct ocfs2_lock_res *lockres,
>>>>>>> +static int __lockres_remove_mask_waiter(struct ocfs2_lock_res *lockres,
>>>>>>>  				      struct ocfs2_mask_waiter *mw)
>>>>>>>  {
>>>>>>> -	unsigned long flags;
>>>>>>>  	int ret = 0;
>>>>>>>  
>>>>>>> -	spin_lock_irqsave(&lockres->l_lock, flags);
>>>>>>> +	assert_spin_locked(&lockres->l_lock);
>>>>>>>  	if (!list_empty(&mw->mw_item)) {
>>>>>>>  		if ((lockres->l_flags & mw->mw_mask) != mw->mw_goal)
>>>>>>>  			ret = -EBUSY;
>>>>>>> @@ -1338,6 +1342,18 @@ static int lockres_remove_mask_waiter(struct ocfs2_lock_res *lockres,
>>>>>>>  		list_del_init(&mw->mw_item);
>>>>>>>  		init_completion(&mw->mw_complete);
>>>>>>>  	}
>>>>>>> +
>>>>>>> +	return ret;
>>>>>>> +}
>>>>>>> +
>>>>>>> +static int lockres_remove_mask_waiter(struct ocfs2_lock_res *lockres,
>>>>>>> +				      struct ocfs2_mask_waiter *mw)
>>>>>>> +{
>>>>>>> +	unsigned long flags;
>>>>>>> +	int ret = 0;
>>>>>>> +
>>>>>>> +	spin_lock_irqsave(&lockres->l_lock, flags);
>>>>>>> +	ret = __lockres_remove_mask_waiter(lockres, mw);
>>>>>>>  	spin_unlock_irqrestore(&lockres->l_lock, flags);
>>>>>>>  
>>>>>>>  	return ret;
>>>>>>> @@ -1373,6 +1389,7 @@ static int __ocfs2_cluster_lock(struct ocfs2_super *osb,
>>>>>>>  	unsigned long flags;
>>>>>>>  	unsigned int gen;
>>>>>>>  	int noqueue_attempted = 0;
>>>>>>> +	int dlm_locked = 0;
>>>>>>>  
>>>>>>>  	ocfs2_init_mask_waiter(&mw);
>>>>>>>  
>>>>>>> @@ -1481,6 +1498,7 @@ again:
>>>>>>>  			ocfs2_recover_from_dlm_error(lockres, 1);
>>>>>>>  			goto out;
>>>>>>>  		}
>>>>>>> +		dlm_locked = 1;
>>>>>>>  
>>>>>>>  		mlog(0, "lock %s, successful return from ocfs2_dlm_lock\n",
>>>>>>>  		     lockres->l_name);
>>>>>>> @@ -1514,10 +1532,17 @@ out:
>>>>>>>  	if (wait && arg_flags & OCFS2_LOCK_NONBLOCK &&
>>>>>>>  	    mw.mw_mask & (OCFS2_LOCK_BUSY|OCFS2_LOCK_BLOCKED)) {
>>>>>>>  		wait = 0;
>>>>>>> -		if (lockres_remove_mask_waiter(lockres, &mw))
>>>>>>> +		spin_lock_irqsave(&lockres->l_lock, flags);
>>>>>>> +		if (__lockres_remove_mask_waiter(lockres, &mw)) {
>>>>>>> +			if (dlm_locked)
>>>>>>> +				lockres_or_flags(lockres,
>>>>>>> +					OCFS2_LOCK_NONBLOCK_FINISHED);
>>>>>>> +			spin_unlock_irqrestore(&lockres->l_lock, flags);
>>>>>>>  			ret = -EAGAIN;
>>>>>>> -		else
>>>>>>> +		} else {
>>>>>>> +			spin_unlock_irqrestore(&lockres->l_lock, flags);
>>>>>>>  			goto again;
>>>>>>> +		}
>>>>>>>  	}
>>>>>>>  	if (wait) {
>>>>>>>  		ret = ocfs2_wait_for_mask(&mw);
>>>>>>> diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h
>>>>>>> index bbec539..7d6b7d0 100644
>>>>>>> --- a/fs/ocfs2/ocfs2.h
>>>>>>> +++ b/fs/ocfs2/ocfs2.h
>>>>>>> @@ -144,6 +144,12 @@ enum ocfs2_unlock_action {
>>>>>>>  						     * before the upconvert
>>>>>>>  						     * has completed */
>>>>>>>  
>>>>>>> +#define OCFS2_LOCK_NONBLOCK_FINISHED (0x00001000) /* NONBLOCK cluster
>>>>>>> +						   * lock has already
>>>>>>> +						   * returned, do not block
>>>>>>> +						   * dc thread from
>>>>>>> +						   * downconverting */
>>>>>>> +
>>>>>>>  struct ocfs2_lock_res_ops;
>>>>>>>  
>>>>>>>  typedef void (*ocfs2_lock_callback)(int status, unsigned long data);
>>>>>>>
>>>>>>
>>>>>> .
>>>>>>
>>>>>
>>>>>
>>>>
>>>> .
>>>>
>>>
>>>
>>
>> .
>>
> 
>