Re: [patch 1/5] mm: add nofail variants of kmalloc kcalloc andkzalloc

[Dave Chinner]

On Wed, Aug 25, 2010 at 09:48:47PM -0400, Ted Ts'o wrote:
> On Wed, Aug 25, 2010 at 05:30:42PM -0700, David Rientjes wrote:
> > 
> > We certainly hope that nobody will reimplement the same function without 
> > the __deprecated warning, especially for order < PAGE_ALLOC_COSTLY_ORDER 
> > where there's no looping at a higher level.  So perhaps the best 
> > alternative is to implement the same _nofail() functions but do a 
> > WARN_ON(get_order(size) > PAGE_ALLOC_COSTLY_ORDER) instead?
> 
> Yeah, that sounds better.
> 
> > I think it's really sad that the caller can't know what the upper bounds 
> > of its memory requirement are ahead of time or at least be able to 
> > implement a memory freeing function when kmalloc() returns NULL.
> 
> Oh, we can determine an upper bound.  You might just not like it.
> Actually ext3/ext4 shouldn't be as bad as XFS, which Dave estimated to
> be around 400k for a transaction.

For a 4k block size filesystem.

If I use 64k block size directories (which XFS can even on 4k page
size machines), the maximum transaction reservation goes up to at
around 3MB, and that's just for blocks being _modified_. It's not
the limit on the amount of memory that may need to be allocated
during a transaction....

> My guess is that the worst case for
> ext3/ext4 is probably around 256k or so; like XFS, most of the time,
> it would be a lot less. 

Right, it usually is a lot less, but one of the big problems is that
during low memory situations memory reclaim of the metadata page
cache actually causes _more_ memory allocation during tranactions
than otherwise would occur.......

Cheers,

Dave.
-- 
Dave Chinner
david@xxxxxxxxxxxxx
--
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/