On Wed, Jul 09, 2014 at 06:24:53PM +0200, Joerg Roedel wrote: > On Tue, Jul 08, 2014 at 05:59:59PM -0400, j.glisse@xxxxxxxxx wrote: > > From: Jérôme Glisse <jglisse@xxxxxxxxxx> > > > > New code will need to be able to differentiate [...] > > Why? > > > between a regular unmap and an unmap trigger by vmscan in which case > > we want to be as quick as possible. > > Why want you be slower as possible in the other case? > Well i should probably have updated the commit message. Trying to be faster is one side of the coin. The other is to actualy know that the page is considered for reclaim which is useful information in itself. Secondary user of the page might want to mark the page as recently use and move it to the active list. In most case an unmap event is not as critical as a vmscan can be (note that this depend on the current states of memory resources if they are scarce or not). While right now there is no special code inside hmm for offering a fast path this is definitly something that is envision on some hardware. Again it is all about trying to take best course of action depending on context and the more informations we have about current context the easier it is to properly choose. Cheers, Jérôme -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>