I just did some debugging on PAE SMP dom0 with Xen on a machine with>4GB memory. As far as we tested, PAE SMP dom0 hangs on machines with >4GB (latest, changeset: 10363:61a81fab2a02).Now it works fines with the patch below. I don''t think this is the final patch, but it''s a hint for debugging (for Keir, especially ;-). If I back out the patch, what I see is that the second level page table for mapping Xen is not valid, i.e. zero clean (L3 looks okay). So debug Xen works fine (i.e. the one built by "make debug=y"). ---- diff -r 61a81fab2a02 xen/arch/x86/mm.c --- a/xen/arch/x86/mm.c Fri Jun 16 14:34:21 2006 +++ b/xen/arch/x86/mm.c Fri Jun 16 16:44:39 2006 @@ -258,11 +258,6 @@ } #if defined(CONFIG_X86_PAE) - -#ifdef NDEBUG -/* Only PDPTs above 4GB boundary need to be shadowed in low memory. */ -#define l3tab_needs_shadow(mfn) (mfn >= 0x100000) -#else /* * In debug builds we aggressively shadow PDPTs to exercise code paths. * We cannot safely shadow the idle page table, nor shadow-mode page tables @@ -272,7 +267,6 @@ ((((mfn << PAGE_SHIFT) != __pa(idle_pg_table)) && \ (page_get_owner(mfn_to_page(mfn)) != NULL)) || \ (mfn >= 0x100000)) -#endif static l1_pgentry_t *fix_pae_highmem_pl1e; Jun --- Intel Open Source Technology Center _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On 17 Jun 2006, at 00:43, Nakajima, Jun wrote:> I just did some debugging on PAE SMP dom0 with Xen on a machine with >> 4GB memory. As far as we tested, PAE SMP dom0 hangs on machines with >> 4GB (latest, changeset: 10363:61a81fab2a02). > > Now it works fines with the patch below. I don''t think this is the > final > patch, but it''s a hint for debugging (for Keir, especially ;-). If I > back out the patch, what I see is that the second level page table for > mapping Xen is not valid, i.e. zero clean (L3 looks okay). So debug Xen > works fine (i.e. the one built by "make debug=y").So you see this problem booting domain0? No shadow pagetables or anything? -- Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On 17 Jun 2006, at 00:43, Nakajima, Jun wrote:> Now it works fines with the patch below. I don''t think this is the > final > patch, but it''s a hint for debugging (for Keir, especially ;-). If I > back out the patch, what I see is that the second level page table for > mapping Xen is not valid, i.e. zero clean (L3 looks okay). So debug Xen > works fine (i.e. the one built by "make debug=y").That''s weird. Is the L2 page itself owned by the right domain and type l2_page_table? What is CR3 at that point (does it point at one of the lowmem caches, or directly at a guest page)? Etc... If you''ve got a reliable reproduction of this problem then I expect we can gather enough tracing to work out what''s going on. -- Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel