Re: Xen 4.0.0-rc7 problem/hang with vt-d DMAR parsing

[Weidong Han <weidong.han@intel.com>]

Jan Beulich wrote:
>>>> Weidong Han <weidong.han@intel.com> 24.03.10 10:02 >>>
>>>>         
>> it cannot check entry_header->length < sizeof(struct acpi_table_XXX), 
>> which is not the actual size in acpi table.
>>     
>
> I don't follow here: Minimally checking against
> sizeof(struct acpi_dmar_entry_header) should be possible. But I can't
> even see why checking for sizeof(struct acpi_table_XXX) in the
> individual case statements can't be done.
>
> Jan
>   
Re-checked the code. You're right. Updated the patch to check with 
sizeof(struct acpi_table_XXX).

Idea-by: Jan Beulich <jbeulich@novell.com <mailto:jbeulich@novell.com>>
Signed-off-by: Weidong Han <weidong.han@intel.com>

diff -r a4eac162dcb9 xen/drivers/passthrough/vtd/dmar.c
--- a/xen/drivers/passthrough/vtd/dmar.c    Thu Mar 25 01:05:03 2010 +0800
+++ b/xen/drivers/passthrough/vtd/dmar.c    Thu Mar 25 03:53:21 2010 +0800
@@ -659,6 +659,23 @@ static int __init acpi_parse_dmar(struct
     while ( ((unsigned long)entry_header) <
             (((unsigned long)dmar) + table->length) )
     {
+        /*
+         * entry_header length should not smaller than size of
+         * any acpi dmar structures. also avoid endless looping
+         * when the lenght is 0 on some bad BIOSs
+         */
+        if ( entry_header->length < sizeof(struct acpi_table_drhd) &&
+             entry_header->length < sizeof(struct acpi_table_rmrr) &&
+             entry_header->length < sizeof(struct acpi_table_atsr) &&
+             entry_header->length < sizeof(struct acpi_table_rhsa) )
+        {
+            dprintk(XENLOG_WARNING VTDPREFIX,
+                    "Invalid entry_header length: 0x%x\n",
+                    entry_header->length);
+            ret = -EINVAL;
+            break;
+        }
+
         switch ( entry_header->type )
         {
         case ACPI_DMAR_DRHD: