Re: [tip:x86/asm] x86: Introduce GDT_ENTRY_INIT()

["H. Peter Anvin" <hpa@zytor.com>]

On 08/03/2009 08:11 AM, Akinobu Mita wrote:
>>>>
>> OK, I spotted the following error, which certainly would explain a crash on
>> an APM machine:
>>
>>        /* data */
>> -       [GDT_ENTRY_APMBIOS_BASE+2]      = { { { 0x0000ffff, 0x00409200 } }
>> },
>> +       [GDT_ENTRY_APMBIOS_BASE+2]      = GDT_ENTRY_INIT(0x409a, 0, 0xffff),
> 
> Oops, I really should have checked the binary because the compiler should
> generate same code with this patch.
> 

Well, it doesn't even on the best of days, because the last hunk changes
a dynamic initializer to a strange mix between static and dynamic:

-static struct desc_struct bad_bios_desc;
+static struct desc_struct bad_bios_desc = GDT_ENTRY_INIT(0x4092, 0, 0);

 /*
  * At some point we want to use this stack frame pointer to unwind
@@ -476,9 +476,6 @@ void pnpbios_calls_init(union
pnp_bios_install_struct *header)
        pnp_bios_callpoint.offset = header->fields.pm16offset;
        pnp_bios_callpoint.segment = PNP_CS16;

-       bad_bios_desc.a = 0;
-       bad_bios_desc.b = 0x00409200;
-
        set_desc_base(&bad_bios_desc, (unsigned long)__va(0x40UL << 4));
        set_desc_limit(&bad_bios_desc, 4095 - (0x40 << 4));

This is about at strange as it comes; I think we could simply do:

GDT_ENTRY_INIT(0x4092, (unsigned long)__va(0x400UL), 4095-0x400);

Although perhaps the 4095 should really be PAGE_SIZE-1...

	-hpa

-- 
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel.  I don't speak on their behalf.