WARNING - OLD ARCHIVES

This is an archived copy of the Xen.org mailing list, which we have preserved to ensure that existing links to archives are not broken. The live archive, which contains the latest emails, can be found at http://lists.xen.org/
   
 
 
Xen 
 
Home Products Support Community News
 
   
 

xen-ppc-devel

[XenPPC] [xenppc-unstable] Create new vmassist type 'pae_extended_cr3'.

To: xen-ppc-devel@xxxxxxxxxxxxxxxxxxx
Subject: [XenPPC] [xenppc-unstable] Create new vmassist type 'pae_extended_cr3'. Only advertise
From: Xen patchbot-xenppc-unstable <patchbot-xenppc-unstable@xxxxxxxxxxxxxxxxxxx>
Date: Tue, 06 Jun 2006 20:05:34 +0000
Delivery-date: Tue, 06 Jun 2006 13:07:14 -0700
Envelope-to: www-data@xxxxxxxxxxxxxxxxxx
List-help: <mailto:xen-ppc-devel-request@lists.xensource.com?subject=help>
List-id: Xen PPC development <xen-ppc-devel.lists.xensource.com>
List-post: <mailto:xen-ppc-devel@lists.xensource.com>
List-subscribe: <http://lists.xensource.com/cgi-bin/mailman/listinfo/xen-ppc-devel>, <mailto:xen-ppc-devel-request@lists.xensource.com?subject=subscribe>
List-unsubscribe: <http://lists.xensource.com/cgi-bin/mailman/listinfo/xen-ppc-devel>, <mailto:xen-ppc-devel-request@lists.xensource.com?subject=unsubscribe>
Reply-to: xen-ppc-devel@xxxxxxxxxxxxxxxxxxx
Sender: xen-ppc-devel-bounces@xxxxxxxxxxxxxxxxxxx
# HG changeset patch
# User kaf24@xxxxxxxxxxxxxxxxxxxx
# Node ID 8aca850f66adbbc6e42469d27374b685ea9787ed
# Parent  c9e6255cb44a2ec123b7bec9c9447697c747a5f3
Create new vmassist type 'pae_extended_cr3'. Only advertise
pae_pgdir_above_4gb tp guests that have enabled this vmassist.
Control tools ensure all PAE page directories are below 4GB
unless the vmassist is enabled (triggered via an extended-cr3
option in guest Elf header).
Signed-off-by: Keir Fraser <keir@xxxxxxxxxxxxx>
---
 tools/libxc/xc_linux_build.c   |   37 +++++--
 tools/libxc/xc_linux_restore.c |  196 +++++++++++++++++++++++++++++++++++++----
 tools/libxc/xc_linux_save.c    |   25 ++++-
 tools/libxc/xc_load_elf.c      |   11 +-
 tools/libxc/xc_private.c       |   22 ++++
 tools/libxc/xenctrl.h          |    3 
 tools/libxc/xg_private.h       |    3 
 xen/arch/x86/domain_build.c    |    3 
 xen/arch/x86/mm.c              |   15 +++
 xen/common/kernel.c            |    5 -
 xen/common/keyhandler.c        |    5 -
 xen/include/public/xen.h       |   16 +++
 12 files changed, 306 insertions(+), 35 deletions(-)

diff -r c9e6255cb44a -r 8aca850f66ad tools/libxc/xc_linux_build.c
--- a/tools/libxc/xc_linux_build.c      Fri Jun 02 19:14:44 2006 +0100
+++ b/tools/libxc/xc_linux_build.c      Mon Jun 05 10:42:40 2006 +0100
@@ -254,16 +254,32 @@ static int setup_pg_tables_pae(int xc_ha
                                unsigned long *page_array,
                                unsigned long vpt_start,
                                unsigned long vpt_end,
-                               unsigned shadow_mode_enabled)
+                               unsigned shadow_mode_enabled,
+                               unsigned pae_mode)
 {
     l1_pgentry_64_t *vl1tab = NULL, *vl1e = NULL;
     l2_pgentry_64_t *vl2tab = NULL, *vl2e = NULL;
     l3_pgentry_64_t *vl3tab = NULL, *vl3e = NULL;
     uint64_t l1tab, l2tab, l3tab, pl1tab, pl2tab, pl3tab;
-    unsigned long ppt_alloc, count;
+    unsigned long ppt_alloc, count, nmfn;
 
     /* First allocate page for page dir. */
     ppt_alloc = (vpt_start - dsi_v_start) >> PAGE_SHIFT;
+
+    if ( pae_mode == PAEKERN_extended_cr3 )
+    {
+        ctxt->vm_assist |= (1UL << VMASST_TYPE_pae_extended_cr3);
+    }
+    else if ( page_array[ppt_alloc] > 0xfffff )
+    {
+        nmfn = xc_make_page_below_4G(xc_handle, dom, page_array[ppt_alloc]);
+        if ( nmfn == 0 )
+        {
+            fprintf(stderr, "Couldn't get a page below 4GB :-(\n");
+            goto error_out;
+        }
+        page_array[ppt_alloc] = nmfn;
+    }
 
     alloc_pt(l3tab, vl3tab, pl3tab);
     vl3e = &vl3tab[l3_table_offset_pae(dsi_v_start)];
@@ -579,11 +595,11 @@ static int compat_check(int xc_handle, s
     }
 
     if (strstr(xen_caps, "xen-3.0-x86_32p")) {
-        if (!dsi->pae_kernel) {
+        if (dsi->pae_kernel == PAEKERN_no) {
             ERROR("Non PAE-kernel on PAE host.");
             return 0;
         }
-    } else if (dsi->pae_kernel) {
+    } else if (dsi->pae_kernel != PAEKERN_no) {
         ERROR("PAE-kernel on non-PAE host.");
         return 0;
     }
@@ -673,7 +689,8 @@ static int setup_guest(int xc_handle,
 
     for ( i = 0; i < XENFEAT_NR_SUBMAPS; i++ )
     {
-        if ( (supported_features[i]&required_features[i]) != 
required_features[i] )
+        if ( (supported_features[i] & required_features[i]) !=
+             required_features[i] )
         {
             ERROR("Guest kernel does not support a required feature.");
             goto error_out;
@@ -719,7 +736,7 @@ static int setup_guest(int xc_handle,
     (((((_h) + ((1UL<<(_s))-1)) & ~((1UL<<(_s))-1)) - \
     ((_l) & ~((1UL<<(_s))-1))) >> (_s))
 #if defined(__i386__)
-        if ( dsi.pae_kernel )
+        if ( dsi.pae_kernel != PAEKERN_no )
         {
             if ( (1 + /* # L3 */
                   NR(dsi.v_start, v_end, L3_PAGETABLE_SHIFT_PAE) + /* # L2 */
@@ -797,11 +814,11 @@ static int setup_guest(int xc_handle,
 
     /* setup page tables */
 #if defined(__i386__)
-    if (dsi.pae_kernel)
+    if (dsi.pae_kernel != PAEKERN_no)
         rc = setup_pg_tables_pae(xc_handle, dom, ctxt,
                                  dsi.v_start, v_end,
                                  page_array, vpt_start, vpt_end,
-                                 shadow_mode_enabled);
+                                 shadow_mode_enabled, dsi.pae_kernel);
     else
         rc = setup_pg_tables(xc_handle, dom, ctxt,
                              dsi.v_start, v_end,
@@ -824,7 +841,7 @@ static int setup_guest(int xc_handle,
      */
     if ( !shadow_mode_enabled )
     {
-        if ( dsi.pae_kernel )
+        if ( dsi.pae_kernel != PAEKERN_no )
         {
             if ( pin_table(xc_handle, MMUEXT_PIN_L3_TABLE,
                            xen_cr3_to_pfn(ctxt->ctrlreg[3]), dom) )
@@ -958,7 +975,7 @@ static int setup_guest(int xc_handle,
     rc = xc_version(xc_handle, XENVER_version, NULL);
     sprintf(start_info->magic, "xen-%i.%i-x86_%d%s",
             rc >> 16, rc & (0xFFFF), (unsigned int)sizeof(long)*8,
-            dsi.pae_kernel ? "p" : "");
+            (dsi.pae_kernel != PAEKERN_no) ? "p" : "");
     start_info->nr_pages     = nr_pages;
     start_info->shared_info  = guest_shared_info_mfn << PAGE_SHIFT;
     start_info->flags        = flags;
diff -r c9e6255cb44a -r 8aca850f66ad tools/libxc/xc_linux_restore.c
--- a/tools/libxc/xc_linux_restore.c    Fri Jun 02 19:14:44 2006 +0100
+++ b/tools/libxc/xc_linux_restore.c    Mon Jun 05 10:42:40 2006 +0100
@@ -108,7 +108,7 @@ int xc_linux_restore(int xc_handle, int 
                      unsigned int console_evtchn, unsigned long *console_mfn)
 {
     DECLARE_DOM0_OP;
-    int rc = 1, i, n;
+    int rc = 1, i, n, pae_extended_cr3 = 0;
     unsigned long mfn, pfn;
     unsigned int prev_pc, this_pc;
     int verify = 0;
@@ -162,25 +162,83 @@ int xc_linux_restore(int xc_handle, int 
         return 1;
     }
 
-
     if (mlock(&ctxt, sizeof(ctxt))) {
         /* needed for build dom0 op, but might as well do early */
         ERR("Unable to mlock ctxt");
         return 1;
     }
 
-
-    /* Read the saved P2M frame list */
-    if(!(p2m_frame_list = malloc(P2M_FL_SIZE))) {
+    if (!(p2m_frame_list = malloc(P2M_FL_SIZE))) {
         ERR("Couldn't allocate p2m_frame_list array");
         goto out;
     }
 
-    if (!read_exact(io_fd, p2m_frame_list, P2M_FL_SIZE)) {
+    /* Read first entry of P2M list, or extended-info signature (~0UL). */
+    if (!read_exact(io_fd, p2m_frame_list, sizeof(long))) {
+        ERR("read extended-info signature failed");
+        goto out;
+    }
+
+    if (p2m_frame_list[0] == ~0UL) {
+        uint32_t tot_bytes;
+
+        /* Next 4 bytes: total size of following extended info. */
+        if (!read_exact(io_fd, &tot_bytes, sizeof(tot_bytes))) {
+            ERR("read extended-info size failed");
+            goto out;
+        }
+
+        while (tot_bytes) {
+            uint32_t chunk_bytes;
+            char     chunk_sig[4];
+
+            /* 4-character chunk signature + 4-byte remaining chunk size. */
+            if (!read_exact(io_fd, chunk_sig, sizeof(chunk_sig)) ||
+                !read_exact(io_fd, &chunk_bytes, sizeof(chunk_bytes))) {
+                ERR("read extended-info chunk signature failed");
+                goto out;
+            }
+            tot_bytes -= 8;
+
+            /* VCPU context structure? */
+            if (!strncmp(chunk_sig, "vcpu", 4)) {
+                if (!read_exact(io_fd, &ctxt, sizeof(ctxt))) {
+                    ERR("read extended-info vcpu context failed");
+                    goto out;
+                }
+                tot_bytes   -= sizeof(struct vcpu_guest_context);
+                chunk_bytes -= sizeof(struct vcpu_guest_context);
+
+                if (ctxt.vm_assist & (1UL << VMASST_TYPE_pae_extended_cr3))
+                    pae_extended_cr3 = 1;
+            }
+
+            /* Any remaining bytes of this chunk: read and discard. */
+            while (chunk_bytes) {
+                unsigned long sz = chunk_bytes;
+                if ( sz > P2M_FL_SIZE )
+                    sz = P2M_FL_SIZE;
+                if (!read_exact(io_fd, p2m_frame_list, sz)) {
+                    ERR("read-and-discard extended-info chunk bytes failed");
+                    goto out;
+                }
+                chunk_bytes -= sz;
+                tot_bytes   -= sz;
+            }
+        }
+
+        /* Now read the real first entry of P2M list. */
+        if (!read_exact(io_fd, p2m_frame_list, sizeof(long))) {
+            ERR("read first entry of p2m_frame_list failed");
+            goto out;
+        }
+    }
+
+    /* First entry is already read into the p2m array. */
+    if (!read_exact(io_fd, &p2m_frame_list[1], P2M_FL_SIZE - sizeof(long))) {
         ERR("read p2m_frame_list failed");
         goto out;
     }
-
 
     /* We want zeroed memory so use calloc rather than malloc. */
     p2m        = calloc(max_pfn, sizeof(unsigned long));
@@ -331,17 +389,27 @@ int xc_linux_restore(int xc_handle, int 
                 ** A page table page - need to 'uncanonicalize' it, i.e.
                 ** replace all the references to pfns with the corresponding
                 ** mfns for the new domain.
+                **
+                ** On PAE we need to ensure that PGDs are in MFNs < 4G, and
+                ** so we may need to update the p2m after the main loop.
+                ** Hence we defer canonicalization of L1s until then.
                 */
-                if(!uncanonicalize_pagetable(pagetype, page)) {
-                    /*
-                    ** Failing to uncanonicalize a page table can be ok
-                    ** under live migration since the pages type may have
-                    ** changed by now (and we'll get an update later).
-                    */
-                    DPRINTF("PT L%ld race on pfn=%08lx mfn=%08lx\n",
-                            pagetype >> 28, pfn, mfn);
-                    nraces++;
-                    continue;
+                if ((pt_levels != 3) ||
+                    pae_extended_cr3 ||
+                    (pagetype != L1TAB)) {
+
+                    if (!uncanonicalize_pagetable(pagetype, page)) {
+                        /*
+                        ** Failing to uncanonicalize a page table can be ok
+                        ** under live migration since the pages type may have
+                        ** changed by now (and we'll get an update later).
+                        */
+                        DPRINTF("PT L%ld race on pfn=%08lx mfn=%08lx\n",
+                                pagetype >> 28, pfn, mfn);
+                        nraces++;
+                        continue;
+                    }
+
                 }
 
             } else if(pagetype != NOTAB) {
@@ -389,6 +457,100 @@ int xc_linux_restore(int xc_handle, int 
     }
 
     DPRINTF("Received all pages (%d races)\n", nraces);
+
+    if ((pt_levels == 3) && !pae_extended_cr3) {
+
+        /*
+        ** XXX SMH on PAE we need to ensure PGDs are in MFNs < 4G. This
+        ** is a little awkward and involves (a) finding all such PGDs and
+        ** replacing them with 'lowmem' versions; (b) upating the p2m[]
+        ** with the new info; and (c) canonicalizing all the L1s using the
+        ** (potentially updated) p2m[].
+        **
+        ** This is relatively slow (and currently involves two passes through
+        ** the pfn_type[] array), but at least seems to be correct. May wish
+        ** to consider more complex approaches to optimize this later.
+        */
+
+        int j, k;
+
+        /* First pass: find all L3TABs current in > 4G mfns and get new mfns */
+        for (i = 0; i < max_pfn; i++) {
+
+            if (((pfn_type[i] & LTABTYPE_MASK)==L3TAB) && (p2m[i]>0xfffffUL)) {
+
+                unsigned long new_mfn;
+                uint64_t l3ptes[4];
+                uint64_t *l3tab;
+
+                l3tab = (uint64_t *)
+                    xc_map_foreign_range(xc_handle, dom, PAGE_SIZE,
+                                         PROT_READ, p2m[i]);
+
+                for(j = 0; j < 4; j++)
+                    l3ptes[j] = l3tab[j];
+
+                munmap(l3tab, PAGE_SIZE);
+
+                if (!(new_mfn=xc_make_page_below_4G(xc_handle, dom, p2m[i]))) {
+                    ERR("Couldn't get a page below 4GB :-(");
+                    goto out;
+                }
+
+                p2m[i] = new_mfn;
+                if (xc_add_mmu_update(xc_handle, mmu,
+                                      (((unsigned long long)new_mfn)
+                                       << PAGE_SHIFT) |
+                                      MMU_MACHPHYS_UPDATE, i)) {
+                    ERR("Couldn't m2p on PAE root pgdir");
+                    goto out;
+                }
+
+                l3tab = (uint64_t *)
+                    xc_map_foreign_range(xc_handle, dom, PAGE_SIZE,
+                                         PROT_READ | PROT_WRITE, p2m[i]);
+
+                for(j = 0; j < 4; j++)
+                    l3tab[j] = l3ptes[j];
+
+                munmap(l3tab, PAGE_SIZE);
+
+            }
+        }
+
+        /* Second pass: find all L1TABs and uncanonicalize them */
+        j = 0;
+
+        for(i = 0; i < max_pfn; i++) {
+
+            if (((pfn_type[i] & LTABTYPE_MASK)==L1TAB)) {
+                region_mfn[j] = p2m[i];
+                j++;
+            }
+
+            if(i == (max_pfn-1) || j == MAX_BATCH_SIZE) {
+
+                if (!(region_base = xc_map_foreign_batch(
+                          xc_handle, dom, PROT_READ | PROT_WRITE,
+                          region_mfn, j))) {
+                    ERR("map batch failed");
+                    goto out;
+                }
+
+                for(k = 0; k < j; k++) {
+                    if(!uncanonicalize_pagetable(L1TAB,
+                                                 region_base + k*PAGE_SIZE)) {
+                        ERR("failed uncanonicalize pt!");
+                        goto out;
+                    }
+                }
+
+                munmap(region_base, j*PAGE_SIZE);
+                j = 0;
+            }
+        }
+
+    }
 
 
     if (xc_finish_mmu_updates(xc_handle, mmu)) {
diff -r c9e6255cb44a -r 8aca850f66ad tools/libxc/xc_linux_save.c
--- a/tools/libxc/xc_linux_save.c       Fri Jun 02 19:14:44 2006 +0100
+++ b/tools/libxc/xc_linux_save.c       Mon Jun 05 10:42:40 2006 +0100
@@ -818,12 +818,33 @@ int xc_linux_save(int xc_handle, int io_
 
     /* Start writing out the saved-domain record. */
 
-    if(!write_exact(io_fd, &max_pfn, sizeof(unsigned long))) {
+    if (!write_exact(io_fd, &max_pfn, sizeof(unsigned long))) {
         ERR("write: max_pfn");
         goto out;
     }
 
-    if(!write_exact(io_fd, p2m_frame_list, P2M_FL_SIZE)) {
+    /*
+     * Write an extended-info structure to inform the restore code that
+     * a PAE guest understands extended CR3 (PDPTs above 4GB). Turns off
+     * slow paths in the restore code.
+     */
+    if ((pt_levels == 3) &&
+        (ctxt.vm_assist & (1UL << VMASST_TYPE_pae_extended_cr3))) {
+        unsigned long signature = ~0UL;
+        uint32_t tot_sz   = sizeof(struct vcpu_guest_context) + 8;
+        uint32_t chunk_sz = sizeof(struct vcpu_guest_context);
+        char chunk_sig[]  = "vcpu";
+        if (!write_exact(io_fd, &signature, sizeof(signature)) ||
+            !write_exact(io_fd, &tot_sz,    sizeof(tot_sz)) ||
+            !write_exact(io_fd, &chunk_sig, 4) ||
+            !write_exact(io_fd, &chunk_sz,  sizeof(chunk_sz)) ||
+            !write_exact(io_fd, &ctxt,      sizeof(ctxt))) {
+            ERR("write: extended info");
+            goto out;
+        }
+    }
+
+    if (!write_exact(io_fd, p2m_frame_list, P2M_FL_SIZE)) {
         ERR("write: p2m_frame_list");
         goto out;
     }
diff -r c9e6255cb44a -r 8aca850f66ad tools/libxc/xc_load_elf.c
--- a/tools/libxc/xc_load_elf.c Fri Jun 02 19:14:44 2006 +0100
+++ b/tools/libxc/xc_load_elf.c Mon Jun 05 10:42:40 2006 +0100
@@ -122,8 +122,15 @@ static int parseelfimage(const char *ima
             ERROR("Actually saw: '%s'", guestinfo);
             return -EINVAL;
         }
-        if ( (strstr(guestinfo, "PAE=yes") != NULL) )
-            dsi->pae_kernel = 1;
+
+        dsi->pae_kernel = PAEKERN_no;
+        p = strstr(guestinfo, "PAE=yes");
+        if ( p != NULL )
+        {
+            dsi->pae_kernel = PAEKERN_yes;
+            if ( !strncmp(p+7, "[extended-cr3]", 14) )
+                dsi->pae_kernel = PAEKERN_extended_cr3;
+        }
 
         break;
     }
diff -r c9e6255cb44a -r 8aca850f66ad tools/libxc/xc_private.c
--- a/tools/libxc/xc_private.c  Fri Jun 02 19:14:44 2006 +0100
+++ b/tools/libxc/xc_private.c  Mon Jun 05 10:42:40 2006 +0100
@@ -430,6 +430,28 @@ int xc_version(int xc_handle, int cmd, v
     return rc;
 }
 
+unsigned long xc_make_page_below_4G(
+    int xc_handle, uint32_t domid, unsigned long mfn)
+{
+    unsigned long new_mfn;
+
+    if ( xc_domain_memory_decrease_reservation(
+        xc_handle, domid, 1, 0, &mfn) != 0 )
+    {
+        fprintf(stderr,"xc_make_page_below_4G decrease failed. mfn=%lx\n",mfn);
+        return 0;
+    }
+
+    if ( xc_domain_memory_increase_reservation(
+        xc_handle, domid, 1, 0, 32, &new_mfn) != 0 )
+    {
+        fprintf(stderr,"xc_make_page_below_4G increase failed. mfn=%lx\n",mfn);
+        return 0;
+    }
+
+    return new_mfn;
+}
+
 /*
  * Local variables:
  * mode: C
diff -r c9e6255cb44a -r 8aca850f66ad tools/libxc/xenctrl.h
--- a/tools/libxc/xenctrl.h     Fri Jun 02 19:14:44 2006 +0100
+++ b/tools/libxc/xenctrl.h     Mon Jun 05 10:42:40 2006 +0100
@@ -453,6 +453,9 @@ int xc_domain_iomem_permission(int xc_ha
                                unsigned long nr_mfns,
                                uint8_t allow_access);
 
+unsigned long xc_make_page_below_4G(int xc_handle, uint32_t domid,
+                                    unsigned long mfn);
+
 typedef dom0_perfc_desc_t xc_perfc_desc_t;
 /* IMPORTANT: The caller is responsible for mlock()'ing the @desc array. */
 int xc_perfc_control(int xc_handle,
diff -r c9e6255cb44a -r 8aca850f66ad tools/libxc/xg_private.h
--- a/tools/libxc/xg_private.h  Fri Jun 02 19:14:44 2006 +0100
+++ b/tools/libxc/xg_private.h  Mon Jun 05 10:42:40 2006 +0100
@@ -156,6 +156,9 @@ struct domain_setup_info
 
     unsigned long elf_paddr_offset;
 
+#define PAEKERN_no           0
+#define PAEKERN_yes          1
+#define PAEKERN_extended_cr3 2
     unsigned int  pae_kernel;
 
     unsigned int  load_symtab;
diff -r c9e6255cb44a -r 8aca850f66ad xen/arch/x86/domain_build.c
--- a/xen/arch/x86/domain_build.c       Fri Jun 02 19:14:44 2006 +0100
+++ b/xen/arch/x86/domain_build.c       Mon Jun 05 10:42:40 2006 +0100
@@ -301,6 +301,9 @@ int construct_dom0(struct domain *d,
                xen_pae ? "yes" : "no", dom0_pae ? "yes" : "no");
         return -EINVAL;
     }
+
+    if ( xen_pae && !!strstr(dsi.xen_section_string, "PAE=yes[extended-cr3]") )
+        set_bit(VMASST_TYPE_pae_extended_cr3, &d->vm_assist);
 
     if ( (p = strstr(dsi.xen_section_string, "FEATURES=")) != NULL )
     {
diff -r c9e6255cb44a -r 8aca850f66ad xen/arch/x86/mm.c
--- a/xen/arch/x86/mm.c Fri Jun 02 19:14:44 2006 +0100
+++ b/xen/arch/x86/mm.c Mon Jun 05 10:42:40 2006 +0100
@@ -996,6 +996,21 @@ static int alloc_l3_table(struct page_in
     int            i;
 
     ASSERT(!shadow_mode_refcounts(d));
+
+#ifdef CONFIG_X86_PAE
+    /*
+     * PAE pgdirs above 4GB are unacceptable if the guest does not understand
+     * the weird 'extended cr3' format for dealing with high-order address
+     * bits. We cut some slack for control tools (before vcpu0 is initialised).
+     */
+    if ( (pfn >= 0x100000) &&
+         unlikely(!VM_ASSIST(d, VMASST_TYPE_pae_extended_cr3)) &&
+         d->vcpu[0] && test_bit(_VCPUF_initialised, &d->vcpu[0]->vcpu_flags) )
+    {
+        MEM_LOG("PAE pgd must be below 4GB (0x%lx >= 0x100000)", pfn);
+        return 0;
+    }
+#endif
 
     pl3e = map_domain_page(pfn);
     for ( i = 0; i < L3_PAGETABLE_ENTRIES; i++ )
diff -r c9e6255cb44a -r 8aca850f66ad xen/common/kernel.c
--- a/xen/common/kernel.c       Fri Jun 02 19:14:44 2006 +0100
+++ b/xen/common/kernel.c       Mon Jun 05 10:42:40 2006 +0100
@@ -184,6 +184,7 @@ long do_xen_version(int cmd, XEN_GUEST_H
     case XENVER_get_features:
     {
         xen_feature_info_t fi;
+        struct domain *d = current->domain;
 
         if ( copy_from_guest(&fi, arg, 1) )
             return -EFAULT;
@@ -191,7 +192,9 @@ long do_xen_version(int cmd, XEN_GUEST_H
         switch ( fi.submap_idx )
         {
         case 0:
-            fi.submap = (1U << XENFEAT_pae_pgdir_above_4gb);
+            fi.submap = 0;
+            if ( VM_ASSIST(d, VMASST_TYPE_pae_extended_cr3) )
+                fi.submap |= (1U << XENFEAT_pae_pgdir_above_4gb);
             if ( shadow_mode_translate(current->domain) )
                 fi.submap |= 
                     (1U << XENFEAT_writable_page_tables) |
diff -r c9e6255cb44a -r 8aca850f66ad xen/common/keyhandler.c
--- a/xen/common/keyhandler.c   Fri Jun 02 19:14:44 2006 +0100
+++ b/xen/common/keyhandler.c   Mon Jun 05 10:42:40 2006 +0100
@@ -128,11 +128,12 @@ static void dump_domains(unsigned char k
                d->domain_flags, atomic_read(&d->refcnt),
                d->tot_pages, d->xenheap_pages, cpuset);
         printk("    handle=%02x%02x%02x%02x-%02x%02x-%02x%02x-"
-               "%02x%02x-%02x%02x%02x%02x%02x%02x\n",
+               "%02x%02x-%02x%02x%02x%02x%02x%02x vm_assist=%08lx\n",
                d->handle[ 0], d->handle[ 1], d->handle[ 2], d->handle[ 3],
                d->handle[ 4], d->handle[ 5], d->handle[ 6], d->handle[ 7],
                d->handle[ 8], d->handle[ 9], d->handle[10], d->handle[11],
-               d->handle[12], d->handle[13], d->handle[14], d->handle[15]);
+               d->handle[12], d->handle[13], d->handle[14], d->handle[15],
+               d->vm_assist);
 
         arch_dump_domain_info(d);
 
diff -r c9e6255cb44a -r 8aca850f66ad xen/include/public/xen.h
--- a/xen/include/public/xen.h  Fri Jun 02 19:14:44 2006 +0100
+++ b/xen/include/public/xen.h  Mon Jun 05 10:42:40 2006 +0100
@@ -234,10 +234,24 @@ DEFINE_XEN_GUEST_HANDLE(mmuext_op_t);
  */
 #define VMASST_CMD_enable                0
 #define VMASST_CMD_disable               1
+
+/* x86/32 guests: simulate full 4GB segment limits. */
 #define VMASST_TYPE_4gb_segments         0
+
+/* x86/32 guests: trap (vector 15) whenever above vmassist is used. */
 #define VMASST_TYPE_4gb_segments_notify  1
+
+/*
+ * x86 guests: support writes to bottom-level PTEs.
+ * NB1. Page-directory entries cannot be written.
+ * NB2. Guest must continue to remove all writable mappings of PTEs.
+ */
 #define VMASST_TYPE_writable_pagetables  2
-#define MAX_VMASST_TYPE 2
+
+/* x86/PAE guests: support PDPTs above 4GB. */
+#define VMASST_TYPE_pae_extended_cr3     3
+
+#define MAX_VMASST_TYPE                  3
 
 #ifndef __ASSEMBLY__
 

_______________________________________________
Xen-ppc-devel mailing list
Xen-ppc-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-ppc-devel

<Prev in Thread] Current Thread [Next in Thread>
  • [XenPPC] [xenppc-unstable] Create new vmassist type 'pae_extended_cr3'. Only advertise, Xen patchbot-xenppc-unstable <=