From: Jeremy Fitzhardinge Subject: Re: [PATCH]: Fix Xen domU boot with batched mprotect Date: Wed, 15 Oct 2008 09:23:53 -0700 Message-ID: <48F61919.2050005@goop.org> References: <48F5CE10.3060403@redhat.com> <48F6274D.76E4.0078.0@novell.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <48F6274D.76E4.0078.0@novell.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Jan Beulich Cc: Chris Lalancette , "xen-devel@lists.xensource.com" , linux-kernel@vger.kernel.org List-Id: xen-devel@lists.xenproject.org Jan Beulich wrote: >>>> Chris Lalancette 15.10.08 13:03 >>> >>>> >> The right thing to do is to use arbitrary_virt_to_machine, so that we can be >> sure we are modifying the right pfn. This unfortunately introduces a >> performance penalty because of a full page-table-walk, but we can avoid that >> penalty for pages in the p2m list by checking if virt_addr_valid is true, and if >> so, just doing the lookup in the p2m table. >> > > Could you explain how virt_addr_valid() can validly be used here? Looking > at its implementation > > #define virt_addr_valid(kaddr) pfn_valid(__pa(kaddr) >> PAGE_SHIFT) > > a kaddr in kmap space (i.e. above high_memory) would return a bogus > physical address, and hence pfn_valid() on the resulting pfn is meaningless. > virt_addr_valid() is supposed to be usable in this circumstace. The comment says "virt_to_page(kaddr) returns a valid pointer if and only if virt_addr_valid(kaddr) returns true", which implies that virt_addr_valid() returns a meaningful result on all addresses - and if not, it should be fixed. > I'd instead simply compare the address in question against high_memory, > and perhaps instead of in arbitrary_virt_to_machine() in > ptep_modify_prot_commit() under an #ifdef CONFIG_HIGHPTE. I suppose, but I don't think there's much cost in making it generally robust. > But > performance-wise, CONFIG_HIGHPTE sucks under Xen anyway, so you'd > better not turn this on in the first place. We may want/need to provide > a means to disable this at run time so the same kernel when run natively > could still make use of it, but without impacting performance under Xen. > That's a secondary issue. What's the source of the performance hit? Just all the extra kmap_atomic operations? J