On Wed 08-01-14 14:19:23, Raghavendra K T wrote:[...]On 01/07/2014 03:43 AM, Andrew Morton wrote:On Mon, 6 Jan 2014 15:51:55 +0530 Raghavendra K T <raghavendra.kt@xxxxxxxxxxxxxxxxxx> wrote:
But having said that I am not able to get an idea of sane implementationSo if we would be happy with just fixing corner cases like this, we might
to solve this readahead failure bug overcoming the anomaly you pointed
:(. hints/ideas.. ?? please let me know.
use total node memory size to detect them, can't we? If total node memory
size is 0, we can use 16 MB (or global number of free pages / 2 if we would
be uneasy with fixed 16 MB limit) as an upperbound...