Re: [PATCH 4/4] f2fs: optimize build_free_nids()

[Jaegeuk Kim]

2013-05-08 (ì), 14:24 +0800, Haicheng Li:
> On Tue, May 07, 2013 at 07:33:59PM +0900, Jaegeuk Kim wrote:
> > 2013-05-06 (ì), 23:15 +0800, Haicheng Li:
> > > diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c
> > > index 1fe3fe2..3136224 100644
> > > --- a/fs/f2fs/node.c
> > > +++ b/fs/f2fs/node.c
> > > @@ -1342,6 +1342,8 @@ static void build_free_nids(struct f2fs_sb_info *sbi)
> > >  		if (nid >= nm_i->max_nid)
> > >  			nid = 0;
> > >  
> > > +		if (nm_i->fcnt > 2 * MAX_FREE_NIDS)
> > > +			break;
> > 
> > Could you explain when this can happen?
> 
> I'm thinking of this possible scenario:
> 
> as we don't hold any spinlock to protect the context, add_free_nid() could be 
> called by other thread anytime, e.g. by the gc_thread_func() in background.

The gc_thread_func() is not a proper example here though, the
buid_free_nids() is covered by nm_i->build_lock, so build_free_nids is
entered only one at a time.
In addtion, build_free_nids starts with checking if (nm_i->fcnt >
NAT_ENTRY_PER_BLOCK) in order not to be conducted repeatedely.

> 
> then nm_i->fcnt could be increased as 2 * MAX_FREE_NIDS while i < FREE_NID_PAGES.
> Anything I misconsidered?

Apart from the correctness of this behavior, I'm not sure why we should
strictly manage this threshold value.
Should we really need to do this?

> 
> > IMO, this is an unnecessary condition check, since the below condition
> > that includes FREE_NID_PAGES already limits the number of free nids.
> > Thanks,
> 
> hmm, the pros is that this check may possibly avoid some (< 4) unnecessary while-loop,
> the cons is that too many checks of (nm_i->fcnt > 2 * MAX_FREE_NIDS)
> would make the code looking messy and fragmentary...
>  
> > >  		if (i++ == FREE_NID_PAGES)
> > >  			break;
> > >  	}
> > 
> > -- 
> > Jaegeuk Kim
> > Samsung
> 
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

-- 
Jaegeuk Kim
Samsung

Attachment: signature.asc
Description: This is a digitally signed message part