From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from bombadil.infradead.org ([198.137.202.133]:37206 "EHLO bombadil.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751741AbeCOTxc (ORCPT ); Thu, 15 Mar 2018 15:53:32 -0400 From: Matthew Wilcox To: Alexander Duyck Cc: linux-mm@vger.kernel.org, netdev@vger.kernel.org, Matthew Wilcox Subject: [RFC 0/2] Shrink page_frag_cache Date: Thu, 15 Mar 2018 12:53:27 -0700 Message-Id: <20180315195329.7787-1-willy@infradead.org> Sender: netdev-owner@vger.kernel.org List-ID: From: Matthew Wilcox I've just learned about the page_frag_cache allocator, and now I want to use it everywhere ;-) But before I start using it in other places, I want to see if it can be improved at all. The pfmemalloc flag is pretty specific to how the network stack uses it (with GFP_ATOMIC), and the pagecnt_bias is tricky to understand. I think we can do better by just using the fields in struct page directly. I don't have a suitable setup for performance testing this code ... Alex, is there any chance you'd have time to give this a spin? Matthew Wilcox (2): mm: Use page->mapping to indicate pfmemalloc page_frag_cache: Store metadata in struct page include/linux/mm.h | 16 ++---- include/linux/mm_types.h | 17 +----- mm/page_alloc.c | 143 ++++++++++++++++++++++++----------------------- net/core/skbuff.c | 4 +- 4 files changed, 82 insertions(+), 98 deletions(-) -- 2.16.2