Re: USBFS Memory allocation Bug

[Mel Gorman <mel@csn.ul.ie>]

On Sun, Mar 14, 2010 at 07:31:14PM -0400, Andrew Morton wrote:
> On Thu, 11 Mar 2010 11:56:22 -0500 (EST) Alan Stern <stern@rowland.harvard.edu> wrote:
> 
> > On Wed, 10 Mar 2010, Mel Gorman wrote:
> > 
> > > > > Is there any means for the driver to take the large request, break it up
> > > > > into multiple smaller requests and submit them one at a time?
> > > > 
> > > > In theory almost anything is possible.  But it would be a big effort
> > > > and not consistent with the way the rest of the driver works.
> > > > 
> > > 
> > > Then about the only other suggestion would be a mempool containing a small
> > > number of largest-possible buffers that is enabled if there is no swap
> > > available.
> > 
> > Considering that this is the first report I have heard about this sort 
> > of problem, and that adding swap space would probably fix it, I'm not 
> > inclined to make any changes.
> 
> Adding swap space is unlikely to help here.  For an order-6 allocation
> the page allocator will go into wtf-youre-kidding-me mode and won't
> even bother trying.
> 

It will try and make the allocation and probably enter direct reclaim
and do a lumpy reclaim for contiguous blocks. What it won't do is retry
as many times as an allocation order <= PAGE_ALLOC_COSTLY_ORDER.

> Asking the allocator for 2^6 physically contiguous pages is terribly
> unreliable and shouldn't be done by any kernel code which wants to be
> useful.
> 

This remains true. With swap, the high-order allocation attempt might succeed
many of the times but there will still be situations where it fails.

-- 
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab