> 8K rsize/wsize NFS needs 16K allocation - because the network code needs to
> reserve some space for the headers, so the actual size of the skbuff data
> area is about 8K + 64bytes. The result is that kmalloc chooses the next
> higher power-of-2 pool = 16K.
i think this happens only because we do not realize that we have missed a
high-order-goal, and low order allocations break up the 'seeds' that might
grow into a high-order page. If low-order allocations are aware of the
number of 'seeds' necessary on a given level to grow a high-level page,
and if the high-order allocation knows that we are working to solve it's
problem and it should just wait a bit, then we would populate higher order
pools even on low-mem boxes.
Of couse we can not prevent swapping on overloaded boxes, but this wasnt
the point.
-- mingo
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.altern.org/andrebalsa/doc/lkml-faq.html