[Ocfs2-devel] [PATCH 1/2] ocfs2 fix o2dlm dlm run purgelist

[Joel Becker]

On Tue, Jun 15, 2010 at 09:43:02PM -0700, Srinivas Eeda wrote:
> There are two problems in dlm_run_purgelist
> 
> 1. If a lockres is found to be in use, dlm_run_purgelist keeps trying to purge
> the same lockres instead of trying the next lockres.
> 
> 2. When a lockres is found unused, dlm_run_purgelist releases lockres spinlock
> before setting DLM_LOCK_RES_DROPPING_REF and calls dlm_purge_lockres.
> spinlock is reacquired but in this window lockres can get reused. This leads
> to BUG.
> 
> This patch modifies dlm_run_purgelist to skip lockres if it's in use and purge
>  next lockres. It also sets DLM_LOCK_RES_DROPPING_REF before releasing the
> lockres spinlock protecting it from getting reused.
> 
> Signed-off-by: Srinivas Eeda <srinivas.eeda at oracle.com>

	I don't really like the way you did this.  You're absolutely
right that we need to hold the spinlock while setting DROPPING REF.  But
there's no need to lift the lockres check logic into run_purge_list.

> @@ -257,15 +224,12 @@ static void dlm_run_purge_list(struct dlm_ctxt *dlm,
>  		 * refs on it -- there's no need to keep the lockres
>  		 * spinlock. */
>  		spin_lock(&lockres->spinlock);
> -		unused = __dlm_lockres_unused(lockres);
> -		spin_unlock(&lockres->spinlock);
> -
> -		if (!unused)
> -			continue;
>  
>  		purge_jiffies = lockres->last_used +
>  			msecs_to_jiffies(DLM_PURGE_INTERVAL_MS);
>  
> +		mlog(0, "purging lockres %.*s\n", lockres->lockname.len,
> +		     lockres->lockname.name);
>  		/* Make sure that we want to be processing this guy at
>  		 * this time. */
>  		if (!purge_now && time_after(purge_jiffies, jiffies)) {

	In fact, I'd move the __dlm_lockres_unused() and
purge_now||time_after() checks into dlm_purge_lockres().  It can return
-EBUSY if the lockres is in use.  It can return -ETIME if purge_now==0
and time_after hits.  Then inside run_purge_list() you just do:

		spin_lock(&lockres->spinlock);
		ret = dlm_purge_lockres(dlm, res, purge_now);
		spin_unlock(&lockres->spinlock);
		if (ret == -EAGAIN)
			break;
		else if (ret == -EBUSY) {
			lockres = list_entry(lockres->next);
			continue;
		else if (ret)
			BUG();

	What about the dlm_lockres_get()?  That's only held while we
drop the dlm spinlock in dlm_purge_lockres(), so you can move it there.
You take the kref only after the _unused() and time_after() checks.
	This actually would make run_purge_list() more readable, not
less.

Joel

-- 

"There are only two ways to live your life. One is as though nothing
 is a miracle. The other is as though everything is a miracle."
        - Albert Einstein

Joel Becker
Principal Software Developer
Oracle
E-mail: joel.becker at oracle.com
Phone: (650) 506-8127