Return-Path: Received: from serv2.oss.ntt.co.jp (serv2.oss.ntt.co.jp [222.151.198.100]) by ozlabs.org (Postfix) with ESMTP id EC17DB6EFF for ; Wed, 12 May 2010 15:59:54 +1000 (EST) Message-ID: <4BEA44B8.7050409@oss.ntt.co.jp> Date: Wed, 12 May 2010 15:03:36 +0900 From: Takuya Yoshikawa MIME-Version: 1.0 To: Marcelo Tosatti Subject: Re: [RFC][PATCH 11/12] KVM: introduce new API for getting/switching dirty bitmaps References: <20100504215645.6448af8f.takuya.yoshikawa@gmail.com> <20100504220821.d68bde57.takuya.yoshikawa@gmail.com> <20100511034329.GB3458@amt.cnet> <4BE8F0F2.60706@oss.ntt.co.jp> <20100511140712.GA7063@amt.cnet> In-Reply-To: <20100511140712.GA7063@amt.cnet> Content-Type: text/plain; charset=ISO-8859-1; format=flowed Cc: linux-arch@vger.kernel.org, arnd@arndb.de, kvm@vger.kernel.org, kvm-ia64@vger.kernel.org, fernando@oss.ntt.co.jp, x86@kernel.org, agraf@suse.de, kvm-ppc@vger.kernel.org, linux-kernel@vger.kernel.org, linuxppc-dev@ozlabs.org, mingo@redhat.com, paulus@samba.org, avi@redhat.com, hpa@zytor.com, tglx@linutronix.de, Takuya Yoshikawa List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , >>> One alternative would be: >>> >>> KVM_SWITCH_DIRTY_LOG passing the address of a bitmap. If the active >>> bitmap was clean, it returns 0, no switch performed. If the active >>> bitmap was dirty, the kernel switches to the new bitmap and returns 1. >>> >>> And the responsability of cleaning the new bitmap could also be left >>> for userspace. >>> >> >> That is a beautiful approach but we can do that only when we give up using >> GET api. >> >> >> I follow you and Avi's advice about that kind of maintenance policy! >> What do you think? > > If you introduce a switch ioctl that frees the bitmap vmalloc'ed by the > current set_memory_region (if its not freed already), after pointing the > memslot to the user supplied one, it should be fine? > You mean switching from vmalloc'ed(not do_mmap'ed) one to user supplied one? It may be possible but makes things really complicated in my view: until some point we use set_bit, and then use set_bit_user, etc. IMO: - # of slots is limited and the size of dirty_bitmap_old pointer is not problematic. - Both user side and kernel side need not allocate buffers every time and once paired buffers are registered, we will reuse the buffers until user side orders to stop logging. - We have a tiny advantage if we need not copy_from_user to get a bitmap address for switch ioctl. => So I think having two __user bitmaps is not a bad thing.