Re: [Qemu-devel] Virtual Machine Generation ID

[Ben Warren]

> On Jan 18, 2017, at 4:02 PM, Ben Warren <address@hidden> wrote:
> 
> Hi Michael,
>> On Jan 17, 2017, at 9:45 AM, Michael S. Tsirkin <address@hidden> wrote:
>> 
>> On Mon, Jan 16, 2017 at 10:57:42AM -0800, Ben Warren wrote:
>>> I think we have a misunderstanding here.  I’m storing the VM
>>> Generation ID __data__ (a GUID) in a fw_cfg blob, not the address.
>> 
>> Yes, I think I gathered this much from the discussion. This is what
>> I'm saying - don't. Have guest loader reserve guest memory and write the
>> address into a fw cfg blob, and have qemu write the id at that address.
>> This way you can update guest memory at any time.
>> 
> So I’ve gone down the path of creating a writeable fw_cfg blob to hold the 
> VGID address, but it doesn’t seem to be getting updated.
> 
> Here’s the code I’ve added:
> 
> #define VMGENID_FW_CFG_FILE      "etc/vmgenid"
> #define VMGENID_FW_CFG_ADDR_FILE      "etc/vmgenid_addr”
> 
> // Create writeable fw_cfg blob, vas->vgia is a GArray of size 8 and element 
> size 1
> fw_cfg_add_file_callback(s, VMGENID_FW_CFG_ADDR_FILE, NULL, NULL, 
> vms->vgia->data, 8, false);
> 
> // Request BIOS to allocate memory for the read-only DATA file:
> bios_linker_loader_alloc(linker, VMGENID_FW_CFG_FILE, guid, 0,false);
> 
> // Request BIOS to allocate memory for the writeable ADDRESS file:
> bios_linker_loader_alloc(linker, VMGENID_FW_CFG_ADDR_FILE, s->vgia, 0, false);
> 
> // Request BIOS to write the address of the DATA file into the ADDRESS file:
> bios_linker_loader_add_pointer(linker, VMGENID_FW_CFG_ADDR_FILE, 0, 
> sizeof(uint64_t), VMGENID_FW_CFG_FILE, 0);
> 
> I’ve instrumented SeaBIOS and see the requests being made and memcpy to the 
> file happening, but don’t see any changes in QEMU in the memory pointed to by 
> VMGENID_FW_CFG_ADDR_FILE.  Is this how writeable fw_cfg is supposed to work?
> 
Ed explained it to me and I think I get it now.  I realize that you and Igor 
have explained this throughout the e-mail chain, but I’m a little bit slower.

Anyway, is this understanding correct?  BIOS is in fact patching fw_cfg data, 
but it’s in a copy of the fw_cfg file that is in guest space.  In order to get 
this to work we would need to add a new command to the linker/loader protocol 
to write back changes to QEMU after patching, and of course implement the 
change in BIOS and UEFI projects.
> thanks,
> Ben
> 
>> -- 
>> MST
>

smime.p7s
Description: S/MIME cryptographic signature