xen/xsm: Fix xsm_mmu_* and xsm_update_va_mapping hooks

This is an attempt to properly fix the hypervisor crash previously described in
http://marc.info/?l=xen-devel&m=128396289707362&w=2

In looking into this issue, I think the proper fix is to move the xsm_mmu_* and xsm_update_va_mapping hook calls later in the callers, after more validation has been performed and the page_info struct is readily available, and pass the page_info to the hooks. This patch moves the xsm_mmu_normal_update, xsm_mmu_machphys_update and xsm_update_va_mapping hook calls accordingly, and updates their interfaces and hook function implementations. This appears to resolve the crashes for me. However, as I am not overly familiar with this area of code, I'd appreciate any feedback or suggestions for improvement.

Signed-off-by Stephen Smalley <[email protected]>

---

 xen/arch/x86/mm.c     |   29 ++++++++++++++++++-----------
 xen/include/xsm/xsm.h |   28 +++++++++++++++-------------
 xen/xsm/dummy.c       |   11 ++++++-----
 xen/xsm/flask/hooks.c |   41 ++++++++---------------------------------
 4 files changed, 47 insertions(+), 62 deletions(-)

xen/arch/x86/mm.c +18 -11

@@ -3448,9 +3448,6 @@
         {
             p2m_type_t p2mt;
 
-            rc = xsm_mmu_normal_update(d, pg_owner, req.val);
-            if ( rc )
-                break;
             rc = -EINVAL;
 
             req.ptr -= cmd;

@@ -3478,6 +3475,13 @@
                           (unsigned long)(req.ptr & ~PAGE_MASK));
             page = mfn_to_page(mfn);
 
+            rc = xsm_mmu_normal_update(d, req.val, page);
+            if ( rc ) {
+                unmap_domain_page_with_cache(va, &mapcache);
+                put_page(page);
+                break;
+            }
+
             if ( page_lock(page) )
             {
                 switch ( page->u.inuse.type_info & PGT_type_mask )

@@ -3640,10 +3644,6 @@
             mfn = req.ptr >> PAGE_SHIFT;
             gpfn = req.val;
 
-            rc = xsm_mmu_machphys_update(d, mfn);
-            if ( rc )
-                break;
-
             if ( unlikely(!get_page_from_pagenr(mfn, pg_owner)) )
             {
                 MEM_LOG("Could not get page for mach->phys update");

@@ -3658,6 +3658,10 @@
                 break;
             }
 
+            rc = xsm_mmu_machphys_update(d, mfn_to_page(mfn));
+            if ( rc )
+                break;
+
             set_gpfn_from_mfn(mfn, gpfn);
 
             paging_mark_dirty(pg_owner, mfn);

@@ -4272,10 +4276,6 @@
 
     perfc_incr(calls_to_update_va);
 
-    rc = xsm_update_va_mapping(d, pg_owner, val);
-    if ( rc )
-        return rc;
-
     rc = -EINVAL;
     pl1e = guest_map_l1e(v, va, &gl1mfn);
     if ( unlikely(!pl1e || !get_page_from_pagenr(gl1mfn, d)) )

@@ -4295,6 +4295,13 @@
         goto out;
     }
 
+    rc = xsm_update_va_mapping(d, val, gl1pg);
+    if ( rc ) {
+        page_unlock(gl1pg);
+        put_page(gl1pg);
+        goto out;
+    }
+
     rc = mod_l1_entry(pl1e, val, gl1mfn, 0, v, pg_owner);
 
     page_unlock(gl1pg);

xen/include/xsm/xsm.h +15 -13

@@ -136,11 +136,12 @@
     int (*getidletime) (void);
     int (*machine_memory_map) (void);
     int (*domain_memory_map) (struct domain *d);
-    int (*mmu_normal_update) (struct domain *d, struct domain *f, 
-                                                                intpte_t fpte);
-    int (*mmu_machphys_update) (struct domain *d, unsigned long mfn);
-    int (*update_va_mapping) (struct domain *d, struct domain *f, 
-                                                            l1_pgentry_t pte);
+    int (*mmu_normal_update) (struct domain *d,
+			      intpte_t fpte, struct page_info *page);
+    int (*mmu_machphys_update) (struct domain *d, struct page_info *page);
+    int (*update_va_mapping) (struct domain *d,
+			      l1_pgentry_t pte,
+			      struct page_info *page);
     int (*add_to_physmap) (struct domain *d1, struct domain *d2);
     int (*sendtrigger) (struct domain *d);
     int (*test_assign_device) (uint32_t machine_bdf);

@@ -567,21 +568,22 @@
     return xsm_call(domain_memory_map(d));
 }
 
-static inline int xsm_mmu_normal_update (struct domain *d, struct domain *f, 
-                                                                intpte_t fpte)
+static inline int xsm_mmu_normal_update (struct domain *d,
+					 intpte_t fpte, struct page_info *page)
 {
-    return xsm_call(mmu_normal_update(d, f, fpte));
+    return xsm_call(mmu_normal_update(d, fpte, page));
 }
 
-static inline int xsm_mmu_machphys_update (struct domain *d, unsigned long mfn)
+static inline int xsm_mmu_machphys_update (struct domain *d, struct page_info *page)
 {
-    return xsm_call(mmu_machphys_update(d, mfn));
+    return xsm_call(mmu_machphys_update(d, page));
 }
 
-static inline int xsm_update_va_mapping(struct domain *d, struct domain *f, 
-                                                            l1_pgentry_t pte)
+static inline int xsm_update_va_mapping(struct domain *d,
+					l1_pgentry_t pte,
+					struct page_info *page)
 {
-    return xsm_call(update_va_mapping(d, f, pte));
+    return xsm_call(update_va_mapping(d, pte, page));
 }
 
 static inline int xsm_add_to_physmap(struct domain *d1, struct domain *d2)

xen/xsm/dummy.c +6 -5

@@ -405,19 +405,20 @@
     return 0;
 }
 
-static int dummy_mmu_normal_update (struct domain *d, struct domain *f, 
-                                                                intpte_t fpte)
+static int dummy_mmu_normal_update (struct domain *d,
+				    intpte_t fpte, struct page_info *page)
 {
     return 0;
 }
 
-static int dummy_mmu_machphys_update (struct domain *d, unsigned long mfn)
+static int dummy_mmu_machphys_update (struct domain *d, struct page_info *page)
 {
     return 0;
 }
 
-static int dummy_update_va_mapping (struct domain *d, struct domain *f, 
-                                                            l1_pgentry_t pte)
+static int dummy_update_va_mapping (struct domain *d,
+				    l1_pgentry_t pte,
+				    struct page_info *page)
 {
     return 0;
 }

xen/xsm/flask/hooks.c +8 -33

@@ -347,26 +347,6 @@
     return rc;
 }
 
-static int get_mfn_sid(unsigned long mfn, u32 *sid)
-{
-    int rc = 0;
-    struct page_info *page;
-
-    if ( mfn_valid(mfn) )
-    {
-        /*mfn is valid if this is a page that Xen is tracking!*/
-        page = mfn_to_page(mfn);
-        rc = get_page_sid(page, sid);
-    }
-    else
-    {
-        /*Possibly an untracked IO page?*/
-        rc = security_iomem_sid(mfn, sid);
-    }
-
-    return rc;    
-}
-
 static int flask_memory_adjust_reservation(struct domain *d1, struct domain *d2)
 {
     return domain_has_perm(d1, d2, SECCLASS_MMU, MMU__ADJUST);

@@ -1006,12 +986,11 @@
     return domain_has_perm(current->domain, d, SECCLASS_MMU, MMU__MEMORYMAP);
 }
 
-static int flask_mmu_normal_update(struct domain *d, struct domain *f, 
-                                   intpte_t fpte)
+static int flask_mmu_normal_update(struct domain *d, 
+                                   intpte_t fpte, struct page_info *page)
 {
     int rc = 0;
     u32 map_perms = MMU__MAP_READ;
-    unsigned long fmfn;
     struct domain_security_struct *dsec;
     u32 fsid;
 

@@ -1020,42 +999,38 @@
     if ( l1e_get_flags(l1e_from_intpte(fpte)) & _PAGE_RW )
         map_perms |= MMU__MAP_WRITE;
 
-    fmfn = gmfn_to_mfn(f, l1e_get_pfn(l1e_from_intpte(fpte)));
-
-    rc = get_mfn_sid(fmfn, &fsid);
+    rc = get_page_sid(page, &fsid);
     if ( rc )
         return rc;
 
     return avc_has_perm(dsec->sid, fsid, SECCLASS_MMU, map_perms, NULL);
 }
 
-static int flask_mmu_machphys_update(struct domain *d, unsigned long mfn)
+static int flask_mmu_machphys_update(struct domain *d, struct page_info *page)
 {
     int rc = 0;
     u32 psid;
     struct domain_security_struct *dsec;
     dsec = d->ssid;
 
-    rc = get_mfn_sid(mfn, &psid);
+    rc = get_page_sid(page, &psid);
     if ( rc )
         return rc;
 
     return avc_has_perm(dsec->sid, psid, SECCLASS_MMU, MMU__UPDATEMP, NULL);
 }
 
-static int flask_update_va_mapping(struct domain *d, struct domain *f, 
-                                   l1_pgentry_t pte)
+static int flask_update_va_mapping(struct domain *d,
+                                   l1_pgentry_t pte, struct page_info *page)
 {
     int rc = 0;
     u32 psid;
     u32 map_perms = MMU__MAP_READ;
-    unsigned long mfn;
     struct domain_security_struct *dsec;
 
     dsec = d->ssid;
 
-    mfn = gmfn_to_mfn(f, l1e_get_pfn(pte));        
-    rc = get_mfn_sid(mfn, &psid);
+    rc = get_page_sid(page, &psid);
     if ( rc )
         return rc;
 

Quoted 14 lines · Stephen Smalley

This is an attempt to properly fix the hypervisor crash previously described in
http://marc.info/?l=xen-devel&m=128396289707362&w=2

In looking into this issue, I think the proper fix is to move the xsm_mmu_* and xsm_update_va_mapping hook calls later in the callers, after more validation has been performed and the page_info struct is readily available, and pass the page_info to the hooks. This patch moves the xsm_mmu_normal_update, xsm_mmu_machphys_update and xsm_update_va_mapping hook calls accordingly, and updates their interfaces and hook function implementations. This appears to resolve the crashes for me. However, as I am not overly familiar with this area of code, I'd appreciate any feedback or suggestions for improvement.

Quoted 1 line

Signed-off-by Stephen Smalley <[email protected]>

Zero replies after two postings. Should I retry, abort, or fail? Seriously though - if there is something wrong with this patch, then let me know, and if not, can it get applied sometime?

Quoted 270 lines

---

 xen/arch/x86/mm.c     |   29 ++++++++++++++++++-----------
 xen/include/xsm/xsm.h |   28 +++++++++++++++-------------
 xen/xsm/dummy.c       |   11 ++++++-----
 xen/xsm/flask/hooks.c |   41 ++++++++---------------------------------
 4 files changed, 47 insertions(+), 62 deletions(-)

xen/arch/x86/mm.c +18 -11

@@ -3448,9 +3448,6 @@
         {
             p2m_type_t p2mt;
 
-            rc = xsm_mmu_normal_update(d, pg_owner, req.val);
-            if ( rc )
-                break;
             rc = -EINVAL;
 
             req.ptr -= cmd;

@@ -3478,6 +3475,13 @@
                           (unsigned long)(req.ptr & ~PAGE_MASK));
             page = mfn_to_page(mfn);
 
+            rc = xsm_mmu_normal_update(d, req.val, page);
+            if ( rc ) {
+                unmap_domain_page_with_cache(va, &mapcache);
+                put_page(page);
+                break;
+            }
+
             if ( page_lock(page) )
             {
                 switch ( page->u.inuse.type_info & PGT_type_mask )

@@ -3640,10 +3644,6 @@
             mfn = req.ptr >> PAGE_SHIFT;
             gpfn = req.val;
 
-            rc = xsm_mmu_machphys_update(d, mfn);
-            if ( rc )
-                break;
-
             if ( unlikely(!get_page_from_pagenr(mfn, pg_owner)) )
             {
                 MEM_LOG("Could not get page for mach->phys update");

@@ -3658,6 +3658,10 @@
                 break;
             }
 
+            rc = xsm_mmu_machphys_update(d, mfn_to_page(mfn));
+            if ( rc )
+                break;
+
             set_gpfn_from_mfn(mfn, gpfn);
 
             paging_mark_dirty(pg_owner, mfn);

@@ -4272,10 +4276,6 @@
 
     perfc_incr(calls_to_update_va);
 
-    rc = xsm_update_va_mapping(d, pg_owner, val);
-    if ( rc )
-        return rc;
-
     rc = -EINVAL;
     pl1e = guest_map_l1e(v, va, &gl1mfn);
     if ( unlikely(!pl1e || !get_page_from_pagenr(gl1mfn, d)) )

@@ -4295,6 +4295,13 @@
         goto out;
     }
 
+    rc = xsm_update_va_mapping(d, val, gl1pg);
+    if ( rc ) {
+        page_unlock(gl1pg);
+        put_page(gl1pg);
+        goto out;
+    }
+
     rc = mod_l1_entry(pl1e, val, gl1mfn, 0, v, pg_owner);
 
     page_unlock(gl1pg);

xen/include/xsm/xsm.h +6 -5

@@ -136,11 +136,12 @@
     int (*getidletime) (void);
     int (*machine_memory_map) (void);
     int (*domain_memory_map) (struct domain *d);
-    int (*mmu_normal_update) (struct domain *d, struct domain *f,
-                                                                intpte_t
 fpte);
-    int (*mmu_machphys_update) (struct domain *d, unsigned long mfn);
-    int (*update_va_mapping) (struct domain *d, struct domain *f,
-                                                            l1_pgentry_t
 pte);
+    int (*mmu_normal_update) (struct domain *d,
+         intpte_t fpte, struct page_info *page);
+    int (*mmu_machphys_update) (struct domain *d, struct page_info *page);
+    int (*update_va_mapping) (struct domain *d,
+         l1_pgentry_t pte,
+         struct page_info *page);
     int (*add_to_physmap) (struct domain *d1, struct domain *d2);

     int (*sendtrigger) (struct domain *d);
     int (*test_assign_device) (uint32_t machine_bdf);

xen/include/xsm/xsm.h +9 -8

@@ -567,21 +568,22 @@
     return xsm_call(domain_memory_map(d));
 }
 
-static inline int xsm_mmu_normal_update (struct domain *d, struct domain *f,
-                                                                intpte_t
 fpte)
+static inline int xsm_mmu_normal_update (struct domain *d,
+      intpte_t fpte, struct page_info *page)
 {
-    return xsm_call(mmu_normal_update(d, f, fpte));
+    return xsm_call(mmu_normal_update(d, fpte, page));
 }
 
-static inline int xsm_mmu_machphys_update (struct domain *d, unsigned long
 mfn)
+static inline int xsm_mmu_machphys_update (struct domain *d, struct
 page_info *page)
 {
-    return xsm_call(mmu_machphys_update(d, mfn));
+    return xsm_call(mmu_machphys_update(d, page));
 }
 
-static inline int xsm_update_va_mapping(struct domain *d, struct domain *f,
-                                                            l1_pgentry_t
 pte)
+static inline int xsm_update_va_mapping(struct domain *d,
+     l1_pgentry_t pte,
+     struct page_info *page)
 {
-    return xsm_call(update_va_mapping(d, f, pte));
+    return xsm_call(update_va_mapping(d, pte, page));
 }
 
 static inline int xsm_add_to_physmap(struct domain *d1, struct domain *d2)

xen/xsm/dummy.c +6 -5

@@ -405,19 +405,20 @@
     return 0;
 }
 
-static int dummy_mmu_normal_update (struct domain *d, struct domain *f,
-                                                                intpte_t
 fpte)
+static int dummy_mmu_normal_update (struct domain *d,
+        intpte_t fpte, struct page_info *page)
 {
     return 0;
 }
 
-static int dummy_mmu_machphys_update (struct domain *d, unsigned long mfn)
+static int dummy_mmu_machphys_update (struct domain *d, struct page_info
 *page)
 {
     return 0;
 }
 
-static int dummy_update_va_mapping (struct domain *d, struct domain *f,
-                                                            l1_pgentry_t
 pte)
+static int dummy_update_va_mapping (struct domain *d,
+        l1_pgentry_t pte,
+        struct page_info *page)

 {
     return 0;

}

xen/xsm/flask/hooks.c +0 -20

@@ -347,26 +347,6 @@
     return rc;
 }
 
-static int get_mfn_sid(unsigned long mfn, u32 *sid)
-{
-    int rc = 0;
-    struct page_info *page;
-
-    if ( mfn_valid(mfn) )
-    {
-        /*mfn is valid if this is a page that Xen is tracking!*/
-        page = mfn_to_page(mfn);
-        rc = get_page_sid(page, sid);
-    }
-    else
-    {
-        /*Possibly an untracked IO page?*/
-        rc = security_iomem_sid(mfn, sid);
-    }
-
-    return rc; 
-}
-
 static int flask_memory_adjust_reservation(struct domain *d1, struct domain
 *d2)
 {

     return domain_has_perm(d1, d2, SECCLASS_MMU, MMU__ADJUST);

xen/xsm/flask/hooks.c +8 -13

@@ -1006,12 +986,11 @@
     return domain_has_perm(current->domain, d, SECCLASS_MMU,
 MMU__MEMORYMAP);
 }
 
-static int flask_mmu_normal_update(struct domain *d, struct domain *f,
-                                   intpte_t fpte)
+static int flask_mmu_normal_update(struct domain *d,
+                                   intpte_t fpte, struct page_info *page)
 {
     int rc = 0;
     u32 map_perms = MMU__MAP_READ;
-    unsigned long fmfn;
     struct domain_security_struct *dsec;
     u32 fsid;

@@ -1020,42 +999,38 @@
     if ( l1e_get_flags(l1e_from_intpte(fpte)) & _PAGE_RW )
         map_perms |= MMU__MAP_WRITE;
 
-    fmfn = gmfn_to_mfn(f, l1e_get_pfn(l1e_from_intpte(fpte)));
-
-    rc = get_mfn_sid(fmfn, &fsid);
+    rc = get_page_sid(page, &fsid);
     if ( rc )
         return rc;
 
     return avc_has_perm(dsec->sid, fsid, SECCLASS_MMU, map_perms, NULL);
 }
 
-static int flask_mmu_machphys_update(struct domain *d, unsigned long mfn)
+static int flask_mmu_machphys_update(struct domain *d, struct page_info
 *page)
 {
     int rc = 0;
     u32 psid;
     struct domain_security_struct *dsec;
     dsec = d->ssid;
 
-    rc = get_mfn_sid(mfn, &psid);
+    rc = get_page_sid(page, &psid);
     if ( rc )
         return rc;
 
     return avc_has_perm(dsec->sid, psid, SECCLASS_MMU, MMU__UPDATEMP, NULL);
 }
 
-static int flask_update_va_mapping(struct domain *d, struct domain *f,
-                                   l1_pgentry_t pte)
+static int flask_update_va_mapping(struct domain *d,
+                                   l1_pgentry_t pte, struct page_info *page)
 {
     int rc = 0;
     u32 psid;
     u32 map_perms = MMU__MAP_READ;
-    unsigned long mfn;
     struct domain_security_struct *dsec;
 
     dsec = d->ssid;
 
-    mfn = gmfn_to_mfn(f, l1e_get_pfn(pte));
-    rc = get_mfn_sid(mfn, &psid);
+    rc = get_page_sid(page, &psid);
     if ( rc )
         return rc;

Quoted 15 lines · Stephen Smalley

Quoted 14 lines · Stephen Smalley

This is an attempt to properly fix the hypervisor crash previously described in
http://marc.info/?l=xen-devel&m=128396289707362&w=2

In looking into this issue, I think the proper fix is to move the xsm_mmu_* and xsm_update_va_mapping hook calls later in the callers, after more validation has been performed and the page_info struct is readily available, and pass the page_info to the hooks. This patch moves the xsm_mmu_normal_update, xsm_mmu_machphys_update and xsm_update_va_mapping hook calls accordingly, and updates their interfaces and hook function implementations. This appears to resolve the crashes for me. However, as I am not overly familiar with this area of code, I'd appreciate any feedback or suggestions for improvement.

Quoted 5 lines

Quoted 1 line

Signed-off-by Stephen Smalley <[email protected]>

Zero replies after two postings. Should I retry, abort, or fail? Seriously though - if there is something wrong with this patch, then let me know, and if not, can it get applied sometime?

Hi Stephen,

The patch was posted as RFC with no signed-off-by. The patch looks fine to me as far as I understand it. Provide a Signed-off-by line and I'll append it to the patch description and apply it.

You can send these kinds of patches without the RFC tag by the way, unless you want to give the opportunity for other NSA people to possibly do review via the public list.