Pardon my uninformed newbie question, but why does domain 0 need to map another domain into its own address space? Isn''t that all in the realm of the hypervisor? Isn''t the hypervisor the only entity that needs 4G support? Shouldn''t domain 0 only need to be able to allocate the pages without having to map them somewhere in its own space? -----Original Message----- From: xen-devel-admin@lists.sourceforge.net [mailto:xen-devel-admin@lists.sourceforge.net]On Behalf Of Gregory Newby Sent: Thursday, November 13, 2003 1:33 PM To: Wesley Parish Cc: Steven Hand; Devel Xen Subject: Re: [Xen-devel] Xen + other stuff? On Thu, Nov 13, 2003 at 10:05:57PM +1300, Wesley Parish wrote:> Oh well, I''ll have to try it then. I''m interested in getting a set offully> secure Linuces running a set of RDBMSes, with application servers on topof> those. > > Anyone hazzard a guess as to how much memory I''ll be needing? I''m certainat> least half a GB - but would that be per XenLinux instance?This relates to something I bumped up against: with the current snapshot, I''ve only been able to allocate about 800MB for domain0. I am also able to allocate ~800MB for a virtual domain. This is on a machine with 2GB of physical memory, another 4GB of swap. When I tried to create another 800MB virtual doamin, I got an error. So, it seems that you''re constained by physical memory, as well as no high memory (4GB support) yet in the kernel. Why I can only get ~800MB, rather than a full 2GB, for a particular domain is something the developers are aware of. Bottom line: get enough RAM for each domain (domain0 + virtuals) to share, but the limit will be ~2GB. When high memory support is added, I understand the limit will be 4GB total (since that''s all the domain0 can address). So, for a new system, 4GB would be the max that Xen can effectively utilize in the immediate future. Or maybe I''m wrong - I''m still trying to learn all this stuff. How, or whether, Xen is able to utilize swap space is not clear to me. But my advice is to consider everything to operate in "real" mode (i.e., no swap at all) -- it''s a good way of having enough memory, even if swapping is available.> Wesley Parish > > P.S. I was forgetting - what''s the latest Linux 2.x.x that people havehad> running on Xen?It''s 2.4.22 (patched, of course). -- Greg> On Wed, 12 Nov 2003 21:10, Steven Hand wrote: > > >Has anyone got Xen working with say, SELinux? Or vserver? > > > > We''ve not tried SELinux, but the vserver patch applies cleanly > > to xenolinux and hence you can do ''two levels of virtualisation'' > > aka k vservers on n xenolinux-es on 1 xen. > > > > cheers, > > > > S. > > > > > > ------------------------------------------------------- > > This SF.Net email sponsored by: ApacheCon 2003, > > 16-19 November in Las Vegas. Learn firsthand the latest > > developments in Apache, PHP, Perl, XML, Java, MySQL, > > WebDAV, and more! http://www.apachecon.com/ > > _______________________________________________ > > Xen-devel mailing list > > Xen-devel@lists.sourceforge.net > > https://lists.sourceforge.net/lists/listinfo/xen-devel > > -- > Clinesterton Beademung - in all of love. > Mau e ki, "He aha te mea nui?" > You ask, "What is the most important thing?" > Maku e ki, "He tangata, he tangata, he tangata." > I reply, "It is people, it is people, it is people." > > > ------------------------------------------------------- > This SF.Net email sponsored by: ApacheCon 2003, > 16-19 November in Las Vegas. Learn firsthand the latest > developments in Apache, PHP, Perl, XML, Java, MySQL, > WebDAV, and more! http://www.apachecon.com/ > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/xen-devel------------------------------------------------------- This SF.Net email sponsored by: ApacheCon 2003, 16-19 November in Las Vegas. Learn firsthand the latest developments in Apache, PHP, Perl, XML, Java, MySQL, WebDAV, and more! http://www.apachecon.com/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel ------------------------------------------------------- This SF.Net email sponsored by: ApacheCon 2003, 16-19 November in Las Vegas. Learn firsthand the latest developments in Apache, PHP, Perl, XML, Java, MySQL, WebDAV, and more! http://www.apachecon.com/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> Pardon my uninformed newbie question, but why does domain 0 need to map > another domain into its own address space? Isn''t that all in the realm of > the hypervisor? Isn''t the hypervisor the only entity that needs 4G support? > > Shouldn''t domain 0 only need to be able to allocate the pages without having > to map them somewhere in its own space?Domain 0 is responsible for many ''control'' operations, such as creating and managing other domains. This includes loading the kernel image for a new domain, for which it needs direct access to the new domain''s memory. Anyway, to avoid confusion here is what we currently support: 1. Up to 4GB memory in the whole system. 2. Up to 800MB per Xenolinux instance. That is, if you have 4GB then all that memory /can/ be used, but spread across several Xenolinux instances. An individual Xenolinux can only make use of 800MB. Despite this, the domain-building tools in domain 0 /can/ access all 4GB of memory. I guess this sounds confusing unless you understand the underlying implementation. Just go with it :-) -- Keir ------------------------------------------------------- This SF.Net email sponsored by: ApacheCon 2003, 16-19 November in Las Vegas. Learn firsthand the latest developments in Apache, PHP, Perl, XML, Java, MySQL, WebDAV, and more! http://www.apachecon.com/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
> Sorry to be obtuse, but I am quite a detailed technical guy, and I would > like to understand the limitation at the code level. > > I looked at the code in xi_build.c (is that the right place)? I saw that the > domain 0 code was copying in the kernel, and initrd images in 1 page at a > time (and deallocating the pfn after each page was copied). The only thing > that looked like needed a large chunk of domain 0 virtual address space was > the pfn array? Is that what is behind the 800mb limit? > > Is the 800Mb an 800Mb physical limit? Could the guests have larger virtual > address spaces, and just swap (within the guest)?The limitation is that, without CONFIG_HIGHMEM, Linux requires a permanent mapping for every physical page that it owns. Since the Linux kernel''s virtual address space starts at 0xC000000, and Xen reserves all address space above 0xFC000000, that means that there''s about 0xFC000000-0xC0000000 == 0x3C000000 == 960MB of virtual address space that Xenolinux could possibly use. Some of this is burned on vmalloc and ioremap areas, bringing thee actual amount available for permanent virtual-to-physical mapping down to not much more than 800MB. With CONFIG_HIGHMEM, Linux no longer needs a permanent mapping for every page it owns. Pages in high memory can be temporarily mapped in (just as the domain builder does in user space). We''ll port this config option real soon now, and that will mean that every Xenolinux instance could be allocated anything up to 4GB. -- Keir ------------------------------------------------------- This SF.Net email sponsored by: ApacheCon 2003, 16-19 November in Las Vegas. Learn firsthand the latest developments in Apache, PHP, Perl, XML, Java, MySQL, WebDAV, and more! http://www.apachecon.com/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
On 13 Nov 2003, at 20:26, Keir Fraser wrote:> With CONFIG_HIGHMEM, Linux no longer needs a permanent mapping for > every page it owns. Pages in high memory can be temporarily mapped in > (just as the domain builder does in user space). We''ll port this > config option real soon now, and that will mean that every Xenolinux > instance could be allocated anything up to 4GB. > > -- KeirTo add, currently there are two modes of High Memory support: (1)4G: It uses 2 level page table entry + dynamic mapping so that each (Xeno)Linux instance can use anything up to 4G of physical memory. (2)64G: It uses 3 level PAE page table entry + dynamic mapping so that each (Xeno)Linux instance can use anything up to 64G of physical memory. -- Bin ------------------------------------------------------- This SF.Net email sponsored by: ApacheCon 2003, 16-19 November in Las Vegas. Learn firsthand the latest developments in Apache, PHP, Perl, XML, Java, MySQL, WebDAV, and more! http://www.apachecon.com/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
On 13 Nov 2003, at 21:02, Bin Ren wrote:> To add, currently there are two modes of High Memory support: > > (1)4G: It uses 2 level page table entry + dynamic mapping so that > each (Xeno)Linux instance can use anything up to 4G of physical > memory. > > (2)64G: It uses 3 level PAE page table entry + dynamic mapping > so that each (Xeno)Linux instance can use anything up to 64G > of physical memory. > > > -- BinFYI, you can explore address space possibilities with simple ''C'' code: #include <stdio.h> char data; int main() { char stack; unsigned long avail; avail = 0xffffffff - (long)&stack; printf("Address available above stack %08lx\n", avail); avail = 0xffffffff - (long)&data; printf("Address available above data %08lx\n", avail); avail = 0xffffffff - (long)main; printf("Address available above code %08lx\n", avail); } -- Bin ------------------------------------------------------- This SF.Net email sponsored by: ApacheCon 2003, 16-19 November in Las Vegas. Learn firsthand the latest developments in Apache, PHP, Perl, XML, Java, MySQL, WebDAV, and more! http://www.apachecon.com/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel
Is this a hugh think to support in 1) the hypervisor? 2) domain 0 control functions? Does the hyper visor support PAE today? If the hypervisor did, then even if the domain 0 didn''t support PAE, (and but the non-0 guest domain did), shouldn''t domain 0 be able to construct and initialize the necessary page mappings on behalf of the guest, and then allow the hypervisor to actually stuff them into the hardware? -----Original Message----- From: xen-devel-admin@lists.sourceforge.net [mailto:xen-devel-admin@lists.sourceforge.net]On Behalf Of Bin Ren Sent: Thursday, November 13, 2003 4:03 PM To: Devel Xen Subject: Re: [Xen-devel] Xen + other stuff? On 13 Nov 2003, at 20:26, Keir Fraser wrote:> With CONFIG_HIGHMEM, Linux no longer needs a permanent mapping for > every page it owns. Pages in high memory can be temporarily mapped in > (just as the domain builder does in user space). We''ll port this > config option real soon now, and that will mean that every Xenolinux > instance could be allocated anything up to 4GB. > > -- KeirTo add, currently there are two modes of High Memory support: (1)4G: It uses 2 level page table entry + dynamic mapping so that each (Xeno)Linux instance can use anything up to 4G of physical memory. (2)64G: It uses 3 level PAE page table entry + dynamic mapping so that each (Xeno)Linux instance can use anything up to 64G of physical memory. -- Bin ------------------------------------------------------- This SF.Net email sponsored by: ApacheCon 2003, 16-19 November in Las Vegas. Learn firsthand the latest developments in Apache, PHP, Perl, XML, Java, MySQL, WebDAV, and more! http://www.apachecon.com/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel ------------------------------------------------------- This SF.Net email sponsored by: ApacheCon 2003, 16-19 November in Las Vegas. Learn firsthand the latest developments in Apache, PHP, Perl, XML, Java, MySQL, WebDAV, and more! http://www.apachecon.com/ _______________________________________________ Xen-devel mailing list Xen-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/xen-devel