Re: [PATCH] nfsd4: set right access bmap when initializing lock stateid

[Mi Jinlong <mijinlong@cn.fujitsu.com>]

J. Bruce Fields:
> On Mon, Mar 28, 2011 at 03:15:09PM +0800, Mi Jinlong wrote:
>>  
>> Content-Type: text/plain; charset=ISO-2022-JP
>> Content-Transfer-Encoding: 7bit
> 
> Thanks, Mi Jinlong, the analysis is helpful, but I don't think your fix
> is right.
> 
> I think the problem here is basically that the cleanup on exit from
> nfsd4_lock() may have to deal with a lock stateid that is partially
> initialized, in that everything has been setup except the stuff that's
> done by get_lock_access().

  You are right.

> 
> Maybe something like this??  But I'm not able to test right now.
> 
> --b.
> 
> diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
> index fbde6f7..9e8ef31 100644
> --- a/fs/nfsd/nfs4state.c
> +++ b/fs/nfsd/nfs4state.c
> @@ -397,10 +397,13 @@ static void unhash_generic_stateid(struct nfs4_stateid *stp)
>  
>  static void free_generic_stateid(struct nfs4_stateid *stp)
>  {
> -	int oflag = nfs4_access_bmap_to_omode(stp);
> +	int oflag;
>  
> -	nfs4_file_put_access(stp->st_file, oflag);
> -	put_nfs4_file(stp->st_file);
> +	if (stp->st_access_bmap) {
> +		nfs4_access_bmap_to_omode(stp);

 This line should be 

     oflag = nfs4_access_bmap_to_omode(stp);

 otherwise, uninitialized oflag will be used at the next line.

 After this patch, kernel runs correctly!

-- 
----
thanks
Mi Jinlong

> +		nfs4_file_put_access(stp->st_file, oflag);
> +		put_nfs4_file(stp->st_file);
> +	}
>  	kmem_cache_free(stateid_slab, stp);
>  }
>  
> 
>