Le lundi 03 mai 2010 à 11:19 +0200, Jiri Slaby a écrit :
quoted text
> On 05/03/2010 11:13 AM, Eric Dumazet wrote:
> > /*
> >  * Warning: if size is not a power of two, kmalloc() might
> >  * waste memory because of its requirements.
> >  */
> > void *kvmalloc(size_t size)
> > {
> > 	void *ptr = kmalloc(size, GFP_KERNEL | __GFP_NOWARN);
> 
> This is kind of waste of high-order pages, isn't it? Wouldn't it be
> better to have a watermark where we skip kmalloc completely?

quoted text
> 
> 

There are many drawbacks using kmalloc() instead of vmalloc().

(NUMA comes to mind, if you want a zone spreaded on all available nodes)

So this kvmalloc() is not meant to be a generic replacement of all
vmalloc() users, only selected ones.

In some spots, allocating 8192 bytes in a driver setup for example, it
should be fine. Using a high order page for this is perfect, as long as
we can have a fallback to vmalloc()

grep bnx2_alloc_mem /proc/vmallocinfo
0xf8690000-0xf8692000    8192 bnx2_alloc_mem+0x111/0x6d0 pages=1 vmalloc


--
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/