x86/mm: Plumbing to allow any PTE update to fail with -ERESTART

Switching to shadow mode is performed in tasklet context.  To facilitate this,
we schedule the tasklet, then create a hypercall continuation to allow the
switch to take place.

As a consequence, the x86 mm code needs to cope with an L1e operation being
continuable.  do_mmu{,ext}_op() may no longer assert that a continuation
doesn't happen on the final iteration.

To handle the arguments correctly on continuation, compat_update_va_mapping*()
may no longer call into their non-compat counterparts.  Move the compat
functions into mm.c rather than exporting __do_update_va_mapping() and
{get,put}_pg_owner(), and fix an unsigned long/int inconsistency with
compat_update_va_mapping_otherdomain().

This is part of XSA-273 / CVE-2018-3620.

Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>

xen/arch/x86/mm.c

@@ -613,6 +613,9 @@ static int alloc_segdesc_page(struct page_info *page)
     return i == 512 ? 0 : -EINVAL;
 }
 
+static int _get_page_type(struct page_info *page, unsigned long type,
+                          bool preemptible);
+
 static int get_page_and_type_from_mfn(
     mfn_t mfn, unsigned long type, struct domain *d,
     int partial, int preemptible)
@@ -624,9 +627,7 @@ static int get_page_and_type_from_mfn(
          unlikely(!get_page_from_mfn(mfn, d)) )
         return -EINVAL;
 
-    rc = (preemptible ?
-          get_page_type_preemptible(page, type) :
-          (get_page_type(page, type) ? 0 : -EINVAL));
+    rc = _get_page_type(page, type, preemptible);
 
     if ( unlikely(rc) && partial >= 0 &&
          (!preemptible || page != current->arch.old_guest_table) )
@@ -1456,8 +1457,7 @@ static int create_pae_xen_mappings(struct domain *d, l3_pgentry_t *pl3e)
     return 1;
 }
 
-static int alloc_l2_table(struct page_info *page, unsigned long type,
-                          int preemptible)
+static int alloc_l2_table(struct page_info *page, unsigned long type)
 {
     struct domain *d = page_get_owner(page);
     unsigned long  pfn = mfn_x(page_to_mfn(page));
@@ -1469,8 +1469,7 @@ static int alloc_l2_table(struct page_info *page, unsigned long type,
 
     for ( i = page->nr_validated_ptes; i < L2_PAGETABLE_ENTRIES; i++ )
     {
-        if ( preemptible && i > page->nr_validated_ptes
-             && hypercall_preempt_check() )
+        if ( i > page->nr_validated_ptes && hypercall_preempt_check() )
         {
             page->nr_validated_ptes = i;
             rc = -ERESTART;
@@ -1481,6 +1480,12 @@ static int alloc_l2_table(struct page_info *page, unsigned long type,
              (rc = get_page_from_l2e(pl2e[i], pfn, d)) > 0 )
             continue;
 
+        if ( unlikely(rc == -ERESTART) )
+        {
+            page->nr_validated_ptes = i;
+            break;
+        }
+
         if ( rc < 0 )
         {
             gdprintk(XENLOG_WARNING, "Failure in alloc_l2_table: slot %#x\n", i);
@@ -1763,7 +1768,7 @@ static void free_l1_table(struct page_info *page)
 }
 
 
-static int free_l2_table(struct page_info *page, int preemptible)
+static int free_l2_table(struct page_info *page)
 {
     struct domain *d = page_get_owner(page);
     unsigned long pfn = mfn_x(page_to_mfn(page));
@@ -1777,7 +1782,7 @@ static int free_l2_table(struct page_info *page, int preemptible)
     do {
         if ( is_guest_l2_slot(d, page->u.inuse.type_info, i) &&
              put_page_from_l2e(pl2e[i], pfn) == 0 &&
-             preemptible && i && hypercall_preempt_check() )
+             i && hypercall_preempt_check() )
         {
            page->nr_validated_ptes = i;
            err = -ERESTART;
@@ -2373,7 +2378,8 @@ static int alloc_page_type(struct page_info *page, unsigned long type,
         rc = alloc_l1_table(page);
         break;
     case PGT_l2_page_table:
-        rc = alloc_l2_table(page, type, preemptible);
+        ASSERT(preemptible);
+        rc = alloc_l2_table(page, type);
         break;
     case PGT_l3_page_table:
         ASSERT(preemptible);
@@ -2463,7 +2469,8 @@ int free_page_type(struct page_info *page, unsigned long type,
         rc = 0;
         break;
     case PGT_l2_page_table:
-        rc = free_l2_table(page, preemptible);
+        ASSERT(preemptible);
+        rc = free_l2_table(page);
         break;
     case PGT_l3_page_table:
         ASSERT(preemptible);
@@ -3550,12 +3557,9 @@ long do_mmuext_op(
     }
 
     if ( rc == -ERESTART )
-    {
-        ASSERT(i < count);
         rc = hypercall_create_continuation(
             __HYPERVISOR_mmuext_op, "hihi",
             uops, (count - i) | MMU_UPDATE_PREEMPTED, pdone, foreigndom);
-    }
     else if ( curr->arch.old_guest_table )
     {
         XEN_GUEST_HANDLE_PARAM(void) null;
@@ -3861,12 +3865,9 @@ long do_mmu_update(
     }
 
     if ( rc == -ERESTART )
-    {
-        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_PARAM(void) null;
@@ -4121,7 +4122,13 @@ static int __do_update_va_mapping(
 long do_update_va_mapping(unsigned long va, u64 val64,
                           unsigned long flags)
 {
-    return __do_update_va_mapping(va, val64, flags, current->domain);
+    int rc = __do_update_va_mapping(va, val64, flags, current->domain);
+
+    if ( rc == -ERESTART )
+        rc = hypercall_create_continuation(
+            __HYPERVISOR_update_va_mapping, "lll", va, val64, flags);
+
+    return rc;
 }
 
 long do_update_va_mapping_otherdomain(unsigned long va, u64 val64,
@@ -4138,6 +4145,46 @@ long do_update_va_mapping_otherdomain(unsigned long va, u64 val64,
 
     put_pg_owner(pg_owner);
 
+    if ( rc == -ERESTART )
+        rc = hypercall_create_continuation(
+            __HYPERVISOR_update_va_mapping_otherdomain,
+            "llli", va, val64, flags, domid);
+
+    return rc;
+}
+
+int compat_update_va_mapping(unsigned int va, uint32_t lo, uint32_t hi,
+                             unsigned int flags)
+{
+    int rc = __do_update_va_mapping(va, ((uint64_t)hi << 32) | lo,
+                                    flags, current->domain);
+
+    if ( rc == -ERESTART )
+        rc = hypercall_create_continuation(
+            __HYPERVISOR_update_va_mapping, "iiii", va, lo, hi, flags);
+
+    return rc;
+}
+
+int compat_update_va_mapping_otherdomain(unsigned int va,
+                                         uint32_t lo, uint32_t hi,
+                                         unsigned int flags, domid_t domid)
+{
+    struct domain *pg_owner;
+    int rc;
+
+    if ( (pg_owner = get_pg_owner(domid)) == NULL )
+        return -ESRCH;
+
+    rc = __do_update_va_mapping(va, ((uint64_t)hi << 32) | lo, flags, pg_owner);
+
+    put_pg_owner(pg_owner);
+
+    if ( rc == -ERESTART )
+        rc = hypercall_create_continuation(
+            __HYPERVISOR_update_va_mapping_otherdomain,
+            "iiiii", va, lo, hi, flags, domid);
+
     return rc;
 }
 

xen/arch/x86/x86_64/compat/mm.c

@@ -163,19 +163,6 @@ int compat_arch_memory_op(unsigned long cmd, XEN_GUEST_HANDLE_PARAM(void) arg)
     return rc;
 }
 
-int compat_update_va_mapping(unsigned int va, u32 lo, u32 hi,
-                             unsigned int flags)
-{
-    return do_update_va_mapping(va, lo | ((u64)hi << 32), flags);
-}
-
-int compat_update_va_mapping_otherdomain(unsigned long va, u32 lo, u32 hi,
-                                         unsigned long flags,
-                                         domid_t domid)
-{
-    return do_update_va_mapping_otherdomain(va, lo | ((u64)hi << 32), flags, domid);
-}
-
 DEFINE_XEN_GUEST_HANDLE(mmuext_op_compat_t);
 
 int compat_mmuext_op(XEN_GUEST_HANDLE_PARAM(void) arg,

xen/include/asm-x86/hypercall.h

@@ -165,7 +165,7 @@ extern int compat_update_va_mapping(
     unsigned int va, u32 lo, u32 hi, unsigned int flags);
 
 extern int compat_update_va_mapping_otherdomain(
-    unsigned long va, u32 lo, u32 hi, unsigned long flags, domid_t domid);
+    unsigned int va, u32 lo, u32 hi, unsigned int flags, domid_t domid);
 
 DEFINE_XEN_GUEST_HANDLE(trap_info_compat_t);
 extern int compat_set_trap_table(XEN_GUEST_HANDLE(trap_info_compat_t) traps);