[Xen-changelog] [xen-unstable] x86-64: don't crash Xen upon direct pv guest access to GDT/LDT mapping area

[Xen patchbot-unstable <patchbot-unstable@xxxxxxxxxxxxxxxxxxx>]

# HG changeset patch
# User Keir Fraser <keir@xxxxxxx>
# Date 1291041272 0
# Node ID 5cd9612db2bba51f63ff7897aca4c72cc4b8e8c2
# Parent  aba70e59a90dcb0cacc2c6834a6ce19d03d88601
x86-64: don't crash Xen upon direct pv guest access to GDT/LDT mapping area

handle_gdt_ldt_mapping_fault() is intended to deal with indirect
accesses (i.e. those caused by descriptor loads) to the GDT/LDT
mapping area only. While for 32-bit segment limits indeed prevent the
function being entered for direct accesses (i.e. a #GP fault will be
raised even before the address translation gets done, on 64-bit even
user mode accesses would lead to control reaching the BUG_ON() at the
beginning of that function.

Fortunately the fix is simple: Since the guest kernel runs in ring 3,
any guest direct access will have the "user mode" bit set, whereas
descriptor loads always do the translations to access the actual
descriptors as kernel mode ones.

Signed-off-by: Jan Beulich <jbeulich@xxxxxxxxxx>

Further, relax the BUG_ON() in handle_gdt_ldt_mapping_fault() to a
check-and-bail. This avoids any problems in future, if we don't
execute x86_64 guest kernels in ring 3 (e.g., because we use a
lightweight HVM container).

Signed-off-by: Keir Fraser <keir@xxxxxxx>
---
 xen/arch/x86/traps.c |   12 +++++++++---
 1 files changed, 9 insertions(+), 3 deletions(-)

diff -r aba70e59a90d -r 5cd9612db2bb xen/arch/x86/traps.c
--- a/xen/arch/x86/traps.c      Fri Nov 26 14:25:30 2010 +0000
+++ b/xen/arch/x86/traps.c      Mon Nov 29 14:34:32 2010 +0000
@@ -1064,8 +1064,14 @@ static int handle_gdt_ldt_mapping_fault(
     unsigned int is_ldt_area = (offset >> (GDT_LDT_VCPU_VA_SHIFT-1)) & 1;
     unsigned int vcpu_area   = (offset >> GDT_LDT_VCPU_VA_SHIFT);
 
-    /* Should never fault in another vcpu's area. */
-    BUG_ON(vcpu_area != curr->vcpu_id);
+    /*
+     * If the fault is in another vcpu's area, it cannot be due to
+     * a GDT/LDT descriptor load. Thus we can reasonably exit immediately, and
+     * indeed we have to since map_ldt_shadow_page() works correctly only on
+     * accesses to a vcpu's own area.
+     */
+    if ( vcpu_area != curr->vcpu_id )
+        return 0;
 
     /* Byte offset within the gdt/ldt sub-area. */
     offset &= (1UL << (GDT_LDT_VCPU_VA_SHIFT-1)) - 1UL;
@@ -1236,7 +1242,7 @@ static int fixup_page_fault(unsigned lon
 
     if ( unlikely(IN_HYPERVISOR_RANGE(addr)) )
     {
-        if ( !(regs->error_code & PFEC_reserved_bit) &&
+        if ( !(regs->error_code & (PFEC_user_mode | PFEC_reserved_bit)) &&
              (addr >= GDT_LDT_VIRT_START) && (addr < GDT_LDT_VIRT_END) )
             return handle_gdt_ldt_mapping_fault(
                 addr - GDT_LDT_VIRT_START, regs);

_______________________________________________
Xen-changelog mailing list
Xen-changelog@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-changelog