xen/CVE-2013-1918-xsa45-7-mm-error-paths-preemptible.patch
Olaf Hering b9d38dfc8d - add xorg-x11-util-devel to BuildRequires to get lndir(1)
- remove xen.migrate.tools_notify_restore_to_hangup_during_migration_--abort_if_busy.patch
  It changed migration protocol and upstream wants a different solution

- bnc#802221 - fix xenpaging
  readd xenpaging.qemu.flush-cache.patch

- Upstream patches from Jan
  26891-x86-S3-Fix-cpu-pool-scheduling-after-suspend-resume.patch
  26930-x86-EFI-fix-runtime-call-status-for-compat-mode-Dom0.patch
- Additional fix for bnc#816159
  CVE-2013-1918-xsa45-followup.patch

- bnc#817068 - Xen guest with >1 sr-iov vf won't start
  xen-managed-pci-device.patch

- Update to Xen 4.2.2 c/s 26064
  The following recent security patches are included in the tarball
  CVE-2013-0151-xsa34.patch (bnc#797285)
  CVE-2012-6075-xsa41.patch (bnc#797523)
  CVE-2013-1917-xsa44.patch (bnc#813673)
  CVE-2013-1919-xsa46.patch (bnc#813675)

- Upstream patch from Jan
  26902-x86-EFI-pass-boot-services-variable-info-to-runtime-code.patch 

- bnc#816159 - VUL-0: xen: CVE-2013-1918: XSA-45: Several long
  latency operations are not preemptible
  CVE-2013-1918-xsa45-1-vcpu-destroy-pagetables-preemptible.patch

OBS-URL: https://build.opensuse.org/package/show/Virtualization/xen?expand=0&rev=237
2013-05-07 14:35:00 +00:00

258 lines
8.1 KiB
Diff

x86: make page table handling error paths preemptible
... as they may take significant amounts of time.
This requires cloning the tweaked continuation logic from
do_mmuext_op() to do_mmu_update().
Note that in mod_l[34]_entry() a negative "preemptible" value gets
passed to put_page_from_l[34]e() now, telling the callee to store the
respective page in current->arch.old_guest_table (for a hypercall
continuation to pick up), rather than carrying out the put right away.
This is going to be made a little more explicit by a subsequent cleanup
patch.
This is part of CVE-2013-1918 / XSA-45.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Acked-by: Tim Deegan <tim@xen.org>
Index: xen-4.2.1-testing/xen/arch/x86/mm.c
===================================================================
--- xen-4.2.1-testing.orig/xen/arch/x86/mm.c
+++ xen-4.2.1-testing/xen/arch/x86/mm.c
@@ -1258,7 +1258,16 @@ static int put_page_from_l3e(l3_pgentry_
#endif
if ( unlikely(partial > 0) )
+ {
+ ASSERT(preemptible >= 0);
return __put_page_type(l3e_get_page(l3e), preemptible);
+ }
+
+ if ( preemptible < 0 )
+ {
+ current->arch.old_guest_table = l3e_get_page(l3e);
+ return 0;
+ }
return put_page_and_type_preemptible(l3e_get_page(l3e), preemptible);
}
@@ -1271,7 +1280,17 @@ static int put_page_from_l4e(l4_pgentry_
(l4e_get_pfn(l4e) != pfn) )
{
if ( unlikely(partial > 0) )
+ {
+ ASSERT(preemptible >= 0);
return __put_page_type(l4e_get_page(l4e), preemptible);
+ }
+
+ if ( preemptible < 0 )
+ {
+ current->arch.old_guest_table = l4e_get_page(l4e);
+ return 0;
+ }
+
return put_page_and_type_preemptible(l4e_get_page(l4e), preemptible);
}
return 1;
@@ -1566,12 +1585,17 @@ static int alloc_l3_table(struct page_in
if ( rc < 0 && rc != -EAGAIN && rc != -EINTR )
{
MEM_LOG("Failure in alloc_l3_table: entry %d", i);
+ if ( i )
+ {
+ page->nr_validated_ptes = i;
+ page->partial_pte = 0;
+ current->arch.old_guest_table = page;
+ }
while ( i-- > 0 )
{
if ( !is_guest_l3_slot(i) )
continue;
unadjust_guest_l3e(pl3e[i], d);
- put_page_from_l3e(pl3e[i], pfn, 0, 0);
}
}
@@ -1601,22 +1625,24 @@ static int alloc_l4_table(struct page_in
page->nr_validated_ptes = i;
page->partial_pte = partial ?: 1;
}
- else if ( rc == -EINTR )
+ else if ( rc < 0 )
{
+ if ( rc != -EINTR )
+ MEM_LOG("Failure in alloc_l4_table: entry %d", i);
if ( i )
{
page->nr_validated_ptes = i;
page->partial_pte = 0;
- rc = -EAGAIN;
+ if ( rc == -EINTR )
+ rc = -EAGAIN;
+ else
+ {
+ if ( current->arch.old_guest_table )
+ page->nr_validated_ptes++;
+ current->arch.old_guest_table = page;
+ }
}
}
- else if ( rc < 0 )
- {
- MEM_LOG("Failure in alloc_l4_table: entry %d", i);
- while ( i-- > 0 )
- if ( is_guest_l4_slot(d, i) )
- put_page_from_l4e(pl4e[i], pfn, 0, 0);
- }
if ( rc < 0 )
return rc;
@@ -2064,7 +2090,7 @@ static int mod_l3_entry(l3_pgentry_t *pl
pae_flush_pgd(pfn, pgentry_ptr_to_slot(pl3e), nl3e);
}
- put_page_from_l3e(ol3e, pfn, 0, 0);
+ put_page_from_l3e(ol3e, pfn, 0, -preemptible);
return rc;
}
@@ -2127,7 +2153,7 @@ static int mod_l4_entry(l4_pgentry_t *pl
return -EFAULT;
}
- put_page_from_l4e(ol4e, pfn, 0, 0);
+ put_page_from_l4e(ol4e, pfn, 0, -preemptible);
return rc;
}
@@ -2285,7 +2311,15 @@ static int alloc_page_type(struct page_i
PRtype_info ": caf=%08lx taf=%" PRtype_info,
page_to_mfn(page), get_gpfn_from_mfn(page_to_mfn(page)),
type, page->count_info, page->u.inuse.type_info);
- page->u.inuse.type_info = 0;
+ if ( page != current->arch.old_guest_table )
+ page->u.inuse.type_info = 0;
+ else
+ {
+ ASSERT((page->u.inuse.type_info &
+ (PGT_count_mask | PGT_validated)) == 1);
+ get_page_light(page);
+ page->u.inuse.type_info |= PGT_partial;
+ }
}
else
{
@@ -3235,21 +3269,17 @@ long do_mmuext_op(
}
if ( (rc = xsm_memory_pin_page(d, pg_owner, page)) != 0 )
- {
- put_page_and_type(page);
okay = 0;
- break;
- }
-
- if ( unlikely(test_and_set_bit(_PGT_pinned,
- &page->u.inuse.type_info)) )
+ else if ( unlikely(test_and_set_bit(_PGT_pinned,
+ &page->u.inuse.type_info)) )
{
MEM_LOG("Mfn %lx already pinned", page_to_mfn(page));
- put_page_and_type(page);
okay = 0;
- break;
}
+ if ( unlikely(!okay) )
+ goto pin_drop;
+
/* A page is dirtied when its pin status is set. */
paging_mark_dirty(pg_owner, page_to_mfn(page));
@@ -3263,7 +3293,13 @@ long do_mmuext_op(
&page->u.inuse.type_info));
spin_unlock(&pg_owner->page_alloc_lock);
if ( drop_ref )
- put_page_and_type(page);
+ {
+ pin_drop:
+ if ( type == PGT_l1_page_table )
+ put_page_and_type(page);
+ else
+ curr->arch.old_guest_table = page;
+ }
}
break;
@@ -3669,11 +3705,28 @@ long do_mmu_update(
void *va;
unsigned long gpfn, gmfn, mfn;
struct page_info *page;
- int rc = 0, i = 0;
- unsigned int cmd, done = 0, pt_dom;
- struct vcpu *v = current;
+ unsigned int cmd, i = 0, done = 0, pt_dom;
+ struct vcpu *curr = current, *v = curr;
struct domain *d = v->domain, *pt_owner = d, *pg_owner;
struct domain_mmap_cache mapcache;
+ int rc = put_old_guest_table(curr);
+
+ if ( unlikely(rc) )
+ {
+ if ( likely(rc == -EAGAIN) )
+ rc = hypercall_create_continuation(
+ __HYPERVISOR_mmu_update, "hihi", ureqs, count, pdone,
+ foreigndom);
+ return rc;
+ }
+
+ if ( unlikely(count == MMU_UPDATE_PREEMPTED) &&
+ likely(guest_handle_is_null(ureqs)) )
+ {
+ /* See the curr->arch.old_guest_table related
+ * hypercall_create_continuation() below. */
+ return (int)foreigndom;
+ }
if ( unlikely(count & MMU_UPDATE_PREEMPTED) )
{
@@ -3722,7 +3775,7 @@ long do_mmu_update(
for ( i = 0; i < count; i++ )
{
- if ( hypercall_preempt_check() )
+ if ( curr->arch.old_guest_table || hypercall_preempt_check() )
{
rc = -EAGAIN;
break;
@@ -3903,9 +3956,27 @@ long do_mmu_update(
}
if ( rc == -EAGAIN )
+ {
+ ASSERT(i < count);
rc = hypercall_create_continuation(
__HYPERVISOR_mmu_update, "hihi",
ureqs, (count - i) | MMU_UPDATE_PREEMPTED, pdone, foreigndom);
+ }
+ else if ( curr->arch.old_guest_table )
+ {
+ XEN_GUEST_HANDLE(void) null;
+
+ ASSERT(rc || i == count);
+ set_xen_guest_handle(null, NULL);
+ /*
+ * In order to have a way to communicate the final return value to
+ * our continuation, we pass this in place of "foreigndom", building
+ * on the fact that this argument isn't needed anymore.
+ */
+ rc = hypercall_create_continuation(
+ __HYPERVISOR_mmu_update, "hihi", null,
+ MMU_UPDATE_PREEMPTED, null, rc);
+ }
put_pg_owner(pg_owner);