Re: [PATCH] ext4: Fix up range_end offset in range cyclic writeback

[Shantanu Goel <sgoel01@yahoo.com>]

[Resending as vger didn't like the HTML message so apologies for the duplicate Ted]


Hi Ted,

Thanks for confirming the fix is good.  We noticed that kupdate writebacks were happening at 
irregular intervals and started inspecting the VM writeback code which 
seemed to be doing its job correctly, and were eventually led to the 
ext4 writeback code.  Without the fix, it was sometimes take a very long time for a file to be flushed to disk asynchronously though synchronous flushes worked fine.  It would be nice if the fix could be included in 
3.3 release cycle.


Regards,Shantanu



________________________________
From: Ted Ts'o <tytso@mit.edu>
To: Shantanu Goel <sgoel01@yahoo.com> 
Cc: Kernel <linux-kernel@vger.kernel.org> 
Sent: Monday, January 9, 2012 10:17 AM
Subject: Re: [PATCH] ext4: Fix up range_end offset in range cyclic writeback

On Sat, Jan 07, 2012 at 07:23:32PM -0800, Shantanu Goel wrote:
> 
> 
> The patch below fixes the range_end calculation which should be a
> byte offset and not a page index.  Also, the patch updates the `end'
> index as well when looping again in range cyclic mode.

Hi Shantanu,

This patch looks good; thanks for spotting this!  Can I ask how you
spotted this and whether it was by code inspection because the
incorrect code was causing some serious problem?

Basically, it's late in the merge window, and I'd rather not include
any new patches unless they are fixing a serious problem at this
point.  Otherwise, I'll make sure it gets merged for the next merge
window.

Thanks,

                        - Ted