Message-ID: <4DD855CD.3060901@panasas.com>
Date: Sun, 22 May 2011 03:16:13 +0300
From: Boaz Harrosh <bharrosh@panasas.com>
To: Benny Halevy <bhalevy@panasas.com>, Fred Isaman <iisaman@netapp.com>,
        Trond Myklebust <Trond.Myklebust@netapp.com>,
        NFS list <linux-nfs@vger.kernel.org>
Subject: Re: Smack on the head: Questions about pnfs write path
References: <4DD84E8E.5040508@panasas.com>
In-Reply-To: <4DD84E8E.5040508@panasas.com>
Content-Type: text/plain; charset=UTF-8
Sender: linux-nfs-owner@vger.kernel.org
MIME-Version: 1.0

On 05/22/2011 02:45 AM, Boaz Harrosh wrote:
> 
> Fred git blame points to you:
> 
> What was the meaning of the below code:
> 

OK Smack *ME* on the head

A layout driver must have a .pg_test function else
that's what happens. pNFS only for a single page.
I'd say it should be opposite no?

Boaz

> @@ -1014,18 +1015,22 @@ static int nfs_flush_one(struct nfs_pageio_descriptor *desc)
>  		*pages++ = req->wb_page;
>  	}
>  	req = nfs_list_entry(data->pages.next);
> 	if ((!lseg) && list_is_singular(&data->pages))
>  		lseg = pnfs_update_layout(desc->pg_inode, req->wb_context,
>  					  req_offset(req), desc->pg_count,
>  					  IOMODE_RW);
> 
> I mean why only the list_is_singular() case gets a pnfs_update_layout?
> 
> Because I have a very funny BUG:
> If I do:
> 	dd if=/dev/zero of=/mnt/pnfs/dd1 bs=4k count=1
>   I get a nice pnfs write out.
> But if I do:
> 	dd if=/dev/zero of=/mnt/pnfs/dd1 bs=4k count=2 (or any > 1)
>   I see all IO going to MDS
> 
> Below patch fixes that:
> @@ -1014,18 +1015,22 @@ static int nfs_flush_one(struct nfs_pageio_descriptor *desc)
>  		*pages++ = req->wb_page;
>  	}
>  	req = nfs_list_entry(data->pages.next);
> -	if ((!lseg) && list_is_singular(&data->pages))
> +	if ((!lseg) /*&& list_is_singular(&data->pages)*/)
>  		lseg = pnfs_update_layout(desc->pg_inode, req->wb_context,
>  					  req_offset(req), desc->pg_count,
>  					  IOMODE_RW);
> 
> Which points to what I'm saying all along. pNFS Write path had no testing
> but by me.
> 
> (BTW am still fighting the unbalanced lseg ref)
> 
> Thanks
> Boaz
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html