Skip to contentSkip to conversation

x86/PoD: correct ordering of checks in p2m_pod_zero_check()

PATCH In review Jan Beulich 3 participants 4 messages Apr 7, 2020

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 x86/PoD: correct ordering of checks in p2m_pod_zero_check()

Commit 0537d246f8db ("mm: add 'is_special_page' inline function...") moved the is_special_page() checks first in its respective changes to PoD code. While this is fine for p2m_pod_zero_check_superpage(), the validity of the MFN is inferred in both cases from the p2m_is_ram() check, which therefore also needs to come first in this 2nd instance.

Take the opportunity and address latent UB here as well - transform the MFN into struct page_info * only after having established that this is a valid page.

---
I will admit that this was build tested only. I did observe the crash late yesterday while in the office, but got around to analyzing it only today, where I'm again restricted in what I can reasonably test.

xen/arch/x86/mm/p2m-pod.c +11 -9
@@ -877,23 +877,25 @@ p2m_pod_zero_check(struct p2m_domain *p2
     for ( i = 0; i < count; i++ )
     {
         p2m_access_t a;
-        struct page_info *pg;
 
         mfns[i] = p2m->get_entry(p2m, gfns[i], types + i, &a,
                                  0, NULL, NULL);
-        pg = mfn_to_page(mfns[i]);
 
         /*
          * If this is ram, and not a pagetable or a special page, and
          * probably not mapped elsewhere, map it; otherwise, skip.
          */
-        if ( !is_special_page(pg) && p2m_is_ram(types[i]) &&
-             (pg->count_info & PGC_allocated) &&
-             !(pg->count_info & PGC_page_table) &&
-             ((pg->count_info & PGC_count_mask) <= max_ref) )
-            map[i] = map_domain_page(mfns[i]);
-        else
-            map[i] = NULL;
+        map[i] = NULL;
+        if ( p2m_is_ram(types[i]) )
+        {
+            const struct page_info *pg = mfn_to_page(mfns[i]);
+
+            if ( !is_special_page(pg) &&
+                 (pg->count_info & PGC_allocated) &&
+                 !(pg->count_info & PGC_page_table) &&
+                 ((pg->count_info & PGC_count_mask) <= max_ref) )
+                map[i] = map_domain_page(mfns[i]);
+        }
     }
 
     /*
Reviewed-by on v1 1/1
#1 lore raw
Quoted 14 lines

-----Original Message-----
From: Jan Beulich <[email protected]>
Sent: 07 April 2020 12:08
To: [email protected]
Cc: Andrew Cooper <[email protected]>; Roger Pau Monné <[email protected]>; Wei Liu <[email protected]>; Paul Durrant <[email protected]>
Subject: [PATCH] x86/PoD: correct ordering of checks in p2m_pod_zero_check()

Commit 0537d246f8db ("mm: add 'is_special_page' inline function...") moved the is_special_page() checks first in its respective changes to PoD code. While this is fine for p2m_pod_zero_check_superpage(), the validity of the MFN is inferred in both cases from the p2m_is_ram() check, which therefore also needs to come first in this 2nd instance.

Quoted 5 lines

Take the opportunity and address latent UB here as well - transform the MFN into struct page_info * only after having established that this is a valid page.

...with a suggestion below

Quoted 22 lines

---
I will admit that this was build tested only. I did observe the crash late yesterday while in the office, but got around to analyzing it only today, where I'm again restricted in what I can reasonably test.

xen/arch/x86/mm/p2m-pod.c +0 -3
@@ -877,23 +877,25 @@ p2m_pod_zero_check(struct p2m_domain *p2
     for ( i = 0; i < count; i++ )
     {
         p2m_access_t a;
-        struct page_info *pg;
 
         mfns[i] = p2m->get_entry(p2m, gfns[i], types + i, &a,
                                  0, NULL, NULL);
-        pg = mfn_to_page(mfns[i]);
 
         /*
          * If this is ram, and not a pagetable or a special page, and
          * probably not mapped elsewhere, map it; otherwise, skip.
          */
-        if ( !is_special_page(pg) && p2m_is_ram(types[i]) &&

xen/arch/x86/mm/p2m-pod.c p2m_pod_zero_check(struct p2m_domain *p2 · line 891

Quoted 10 lines
xen/arch/x86/mm/p2m-pod.c +4 -6
-             (pg->count_info & PGC_allocated) &&
-             !(pg->count_info & PGC_page_table) &&
-             ((pg->count_info & PGC_count_mask) <= max_ref) )
-            map[i] = map_domain_page(mfns[i]);
-        else
-            map[i] = NULL;
+        map[i] = NULL;
+        if ( p2m_is_ram(types[i]) )
+        {
+            const struct page_info *pg = mfn_to_page(mfns[i]);

Perhaps have local scope stack variable for count_info too?

Quoted 10 lines
xen/arch/x86/mm/p2m-pod.c +7 -0
+
+            if ( !is_special_page(pg) &&
+                 (pg->count_info & PGC_allocated) &&
+                 !(pg->count_info & PGC_page_table) &&
+                 ((pg->count_info & PGC_count_mask) <= max_ref) )
+                map[i] = map_domain_page(mfns[i]);
+        }
     }
 
     /*
#2 lore raw
Quoted 14 lines · Paul Durrant
Quoted 14 lines

-----Original Message-----
From: Jan Beulich <[email protected]>
Sent: 07 April 2020 12:08
To: [email protected]
Cc: Andrew Cooper <[email protected]>; Roger Pau Monné <[email protected]>; Wei Liu <[email protected]>; Paul Durrant <[email protected]>
Subject: [PATCH] x86/PoD: correct ordering of checks in p2m_pod_zero_check()

Commit 0537d246f8db ("mm: add 'is_special_page' inline function...") moved the is_special_page() checks first in its respective changes to PoD code. While this is fine for p2m_pod_zero_check_superpage(), the validity of the MFN is inferred in both cases from the p2m_is_ram() check, which therefore also needs to come first in this 2nd instance.

Quoted 7 lines
Quoted 5 lines

Take the opportunity and address latent UB here as well - transform the MFN into struct page_info * only after having established that this is a valid page.

Thanks.

Quoted 28 lines

...with a suggestion below

Quoted 26 lines

---
I will admit that this was build tested only. I did observe the crash late yesterday while in the office, but got around to analyzing it only today, where I'm again restricted in what I can reasonably test.

xen/arch/x86/mm/p2m-pod.c +0 -7
@@ -877,23 +877,25 @@ p2m_pod_zero_check(struct p2m_domain *p2
     for ( i = 0; i < count; i++ )
     {
         p2m_access_t a;
-        struct page_info *pg;
 
         mfns[i] = p2m->get_entry(p2m, gfns[i], types + i, &a,
                                  0, NULL, NULL);
-        pg = mfn_to_page(mfns[i]);
 
         /*
          * If this is ram, and not a pagetable or a special page, and
          * probably not mapped elsewhere, map it; otherwise, skip.
          */
-        if ( !is_special_page(pg) && p2m_is_ram(types[i]) &&
-             (pg->count_info & PGC_allocated) &&
-             !(pg->count_info & PGC_page_table) &&
-             ((pg->count_info & PGC_count_mask) <= max_ref) )
-            map[i] = map_domain_page(mfns[i]);
Quoted 8 lines

xen/arch/x86/mm/p2m-pod.c p2m_pod_zero_check(struct p2m_domain *p2 · line 891

Quoted 6 lines
xen/arch/x86/mm/p2m-pod.c +4 -2
-        else
-            map[i] = NULL;
+        map[i] = NULL;
+        if ( p2m_is_ram(types[i]) )
+        {
+            const struct page_info *pg = mfn_to_page(mfns[i]);

Perhaps have local scope stack variable for count_info too?

I'd view this as useful only if ...

Quoted 2 lines

xen/arch/x86/mm/p2m-pod.c p2m_pod_zero_check(struct p2m_domain *p2 · line 893

Quoted 2 lines
xen/arch/x86/mm/p2m-pod.c +2 -0
+
+            if ( !is_special_page(pg) &&

... this could then also be made make use of it.

Jan

Acked-by on v1 1/1
#3 lore raw
In reply to Jan Beulich
Quoted 11 lines · Jan Beulich

Commit 0537d246f8db ("mm: add 'is_special_page' inline function...") moved the is_special_page() checks first in its respective changes to PoD code. While this is fine for p2m_pod_zero_check_superpage(), the validity of the MFN is inferred in both cases from the p2m_is_ram() check, which therefore also needs to come first in this 2nd instance.

Take the opportunity and address latent UB here as well - transform the MFN into struct page_info * only after having established that this is a valid page.