[Xen-devel] [PATCH v3 2/5] xentrace: Memory/Page Mapping support for DOMID_XEN on ARM

[Benjamin Sanda <ben.sanda@xxxxxxxxxxxxxxx>]

Modified ARM arch specific p2m.c and mm.c files to provide support for
xentrace. Case for DOMID_XEN added xenmem_add_to_physmap_one() when
XENMAPSPACE_gmfn_foreign domain requested via get_pg_owner. This
provides correct calls to rcu_lock_domain() when DOMID_XEN is
requested.

Check for DOMID_XEN added to p2m_lookup() which skips PFN to MFN
translation. DOMID_XEN is considered a PV guest on x86 (i.e MFN ==
GFN), but on ARM there is no such concept. Thus requests to DOMID_XEN
on ARM use a MFN address directly and do not need translation from
PFN.

Check added to p2m_lookup() to determine read/write or read-only page
type when requesting DOMID_XEN. This is  done by checking the
u.inuse.type_info parameter of the requested page. The page rw/ro
paddr_t type is then set accordingly.

---
Changed since v2:

  * Combined v2 patches 2 and 8 into one patch for v3. No code changes.

---
Changed since v1:

  * Moved get_pg_owner() from the arch specific mm.c source files into
    the common page_alloc.c source. This was done as get_pg_owner() is
    now needed by both architectures and is essentially identical in
    both.
  * Moved put_pg_owner from the arch specific mm.c source files into
    the common page_alloc.c source to make available to all platforms.
  * Removed DOMID_XEN check from page_to_mfn(page) call in
    xenmem_add_to_physmap_one() per review comment (check considered
    to be in excess of need).
  * Removed forward declare of get_pg_owner from mm.c (arm) as it is
    now in the mm.h common header.
  * Added check to determine page rw/ro type to correctly set the
    paddr_t to p2m_ram_rw or p2m_ram_ro instead of assuming p2m_ram_rw
  * Corrected block comment format to conform to Xen coding standard

Signed-off-by: Benjamin Sanda <ben.sanda@xxxxxxxxxxxxxxx>
---
 xen/arch/arm/mm.c  |  3 ++-
 xen/arch/arm/p2m.c | 35 +++++++++++++++++++++++++++++++----
 2 files changed, 33 insertions(+), 5 deletions(-)

diff --git a/xen/arch/arm/mm.c b/xen/arch/arm/mm.c
index 81f9e2e..19d6c2c 100644
--- a/xen/arch/arm/mm.c
+++ b/xen/arch/arm/mm.c
@@ -1099,7 +1099,8 @@ int xenmem_add_to_physmap_one(
     {
         struct domain *od;
         p2m_type_t p2mt;
-        od = rcu_lock_domain_by_any_id(foreign_domid);
+        od = get_pg_owner(foreign_domid);
+
         if ( od == NULL )
             return -ESRCH;
 
diff --git a/xen/arch/arm/p2m.c b/xen/arch/arm/p2m.c
index a2a9c4b..a99b670 100644
--- a/xen/arch/arm/p2m.c
+++ b/xen/arch/arm/p2m.c
@@ -227,11 +227,38 @@ paddr_t p2m_lookup(struct domain *d, paddr_t paddr, 
p2m_type_t *t)
 {
     paddr_t ret;
     struct p2m_domain *p2m = &d->arch.p2m;
+    struct page_info *page;
+    unsigned long mfn;
 
-    spin_lock(&p2m->lock);
-    ret = __p2m_lookup(d, paddr, t);
-    spin_unlock(&p2m->lock);
-
+    /*
+    * DOMID_XEN is considered a PV guest on x86 (i.e MFN == GFN), but
+    * on ARM there is no such concept. Thus requests to DOMID_XEN
+    * on ARM use a MFN address directly and do not need translation 
+    * from PFN.
+    */
+    if(DOMID_XEN != d->domain_id)
+    {
+        spin_lock(&p2m->lock);
+        ret = __p2m_lookup(d, paddr, t);
+        spin_unlock(&p2m->lock);
+    }
+    else
+    {
+        /* retrieve the page to determine read/write or read only mapping */
+        mfn = paddr >> PAGE_SHIFT;
+        if (mfn_valid(mfn))
+        {
+            page = mfn_to_page(mfn);
+            *t = (page->u.inuse.type_info == PGT_writable_page ? 
+                                p2m_ram_rw : p2m_ram_ro);
+        }
+        else
+        {
+            *t = p2m_invalid;
+        }
+        ret = paddr;
+    }
+    
     return ret;
 }
 
-- 
2.5.0


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel