uvm_pdaemon.c : 1.4.2.8 : swamping disk

Group,

	I haven't visited this BSD code in awhile, but I think
	that the function uvmpd_scan_inactive() is swamping the FS
	when a large number of pages are dirty and not backed
	by swap.

	However, I must be reading the code incorrectly.

	Thus, what limits the number of uvmexp.pgswapout
	pages that are being swap'ed out?

	Shouldn't there be some calc that identifies the number
	of max pages that should be attempted to be cleaned
	within a specified timeframe?

	This normally wouldn't be so bad, but I think it is
	killing the pagein code efficiency.


	Mitchell Erblich
	Sr Software Engineer