From: Jan Beulich <jbeulich@suse.com>
Subject: x86/shadow: shrink struct page_info's shadow_flags to 16 bits

This is to avoid it overlapping the linear_pt_count field needed for PV
domains. Introduce a separate, HVM-only pagetable_dying field to replace
the sole one left in the upper 16 bits.

Note that the accesses to ->shadow_flags in shadow_{pro,de}mote() get
switched to non-atomic, non-bitops operations, as {test,set,clear}_bit()
are not allowed on uint16_t fields and hence their use would have
required ugly casts. This is fine because all updates of the field ought
to occur with the paging lock held, and other updates of it use |= and
&= as well (i.e. using atomic operations here didn't really guard
against potentially racing updates elsewhere).

This is part of XSA-280.

Reported-by: Prgmr.com Security <security@prgmr.com>
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by: Tim Deegan <tim@xen.org>

--- a/xen/arch/x86/mm/shadow/common.c
+++ b/xen/arch/x86/mm/shadow/common.c
@@ -957,10 +957,14 @@ void shadow_promote(struct domain *d, mf
 
     /* Is the page already shadowed? */
     if ( !test_and_set_bit(_PGC_page_table, &page->count_info) )
+    {
         page->shadow_flags = 0;
+        if ( is_hvm_domain(d) )
+            page->pagetable_dying = false;
+    }
 
-    ASSERT(!test_bit(type, &page->shadow_flags));
-    set_bit(type, &page->shadow_flags);
+    ASSERT(!(page->shadow_flags & (1u << type)));
+    page->shadow_flags |= 1u << type;
     TRACE_SHADOW_PATH_FLAG(TRCE_SFLAG_PROMOTE);
 }
 
@@ -969,9 +973,9 @@ void shadow_demote(struct domain *d, mfn
     struct page_info *page = mfn_to_page(gmfn);
 
     ASSERT(test_bit(_PGC_page_table, &page->count_info));
-    ASSERT(test_bit(type, &page->shadow_flags));
+    ASSERT(page->shadow_flags & (1u << type));
 
-    clear_bit(type, &page->shadow_flags);
+    page->shadow_flags &= ~(1u << type);
 
     if ( (page->shadow_flags & SHF_page_type_mask) == 0 )
     {
@@ -2801,7 +2805,7 @@ void sh_remove_shadows(struct domain *d,
     if ( !fast && all && (pg->count_info & PGC_page_table) )
     {
         SHADOW_ERROR("can't find all shadows of mfn %"PRI_mfn" "
-                     "(shadow_flags=%08x)\n",
+                     "(shadow_flags=%04x)\n",
                       mfn_x(gmfn), pg->shadow_flags);
         domain_crash(d);
     }
--- a/xen/arch/x86/mm/shadow/multi.c
+++ b/xen/arch/x86/mm/shadow/multi.c
@@ -3328,8 +3328,8 @@ static int sh_page_fault(struct vcpu *v,
 
     /* Unshadow if we are writing to a toplevel pagetable that is
      * flagged as a dying process, and that is not currently used. */
-    if ( sh_mfn_is_a_page_table(gmfn)
-         && (mfn_to_page(gmfn)->shadow_flags & SHF_pagetable_dying) )
+    if ( sh_mfn_is_a_page_table(gmfn) && is_hvm_domain(d) &&
+         mfn_to_page(gmfn)->pagetable_dying )
     {
         int used = 0;
         struct vcpu *tmp;
@@ -4301,9 +4301,9 @@ int sh_rm_write_access_from_sl1p(struct
     ASSERT(mfn_valid(smfn));
 
     /* Remember if we've been told that this process is being torn down */
-    if ( curr->domain == d )
+    if ( curr->domain == d && is_hvm_domain(d) )
         curr->arch.paging.shadow.pagetable_dying
-            = !!(mfn_to_page(gmfn)->shadow_flags & SHF_pagetable_dying);
+            = mfn_to_page(gmfn)->pagetable_dying;
 
     sp = mfn_to_page(smfn);
 
@@ -4619,10 +4619,10 @@ static void sh_pagetable_dying(struct vc
                    : shadow_hash_lookup(d, mfn_x(gmfn), SH_type_l2_pae_shadow);
         }
 
-        if ( mfn_valid(smfn) )
+        if ( mfn_valid(smfn) && is_hvm_domain(d) )
         {
             gmfn = _mfn(mfn_to_page(smfn)->v.sh.back);
-            mfn_to_page(gmfn)->shadow_flags |= SHF_pagetable_dying;
+            mfn_to_page(gmfn)->pagetable_dying = true;
             shadow_unhook_mappings(d, smfn, 1/* user pages only */);
             flush = 1;
         }
@@ -4659,9 +4659,9 @@ static void sh_pagetable_dying(struct vc
     smfn = shadow_hash_lookup(d, mfn_x(gmfn), SH_type_l4_64_shadow);
 #endif
 
-    if ( mfn_valid(smfn) )
+    if ( mfn_valid(smfn) && is_hvm_domain(d) )
     {
-        mfn_to_page(gmfn)->shadow_flags |= SHF_pagetable_dying;
+        mfn_to_page(gmfn)->pagetable_dying = true;
         shadow_unhook_mappings(d, smfn, 1/* user pages only */);
         /* Now flush the TLB: we removed toplevel mappings. */
         flush_tlb_mask(d->domain_dirty_cpumask);
--- a/xen/arch/x86/mm/shadow/private.h
+++ b/xen/arch/x86/mm/shadow/private.h
@@ -292,8 +292,6 @@ static inline void sh_terminate_list(str
 
 #endif /* (SHADOW_OPTIMIZATIONS & SHOPT_OUT_OF_SYNC) */
 
-#define SHF_pagetable_dying (1u<<31)
-
 static inline int sh_page_has_multiple_shadows(struct page_info *pg)
 {
     u32 shadows;
--- a/xen/include/asm-x86/mm.h
+++ b/xen/include/asm-x86/mm.h
@@ -188,8 +188,15 @@ struct page_info
          * Guest pages with a shadow.  This does not conflict with
          * tlbflush_timestamp since page table pages are explicitly not
          * tracked for TLB-flush avoidance when a guest runs in shadow mode.
+         *
+         * pagetable_dying is used for HVM domains only. The layout here has
+         * to avoid re-use of the space used by linear_pt_count, which (only)
+         * PV guests use.
          */
-        u32 shadow_flags;
+        struct {
+            uint16_t shadow_flags;
+            bool pagetable_dying;
+        };
 
         /* When in use as a shadow, next shadow in this hash chain. */
         __pdx_t next_shadow;