update_va_mapping_otherdomain

update_va_mapping_otherdomain

[Michael R. Hines]

[-- Attachment #1.1: Type: text/plain, Size: 1045 bytes --]


Greetings,

I have a technical question about update_va_mapping_otherdomain.......

I have two components: a C program linked against libxc
and a kernel module which performs the following simple
sequence of events from domain zero:

1. Pause guest (unprivileged)
2. Grabs the PTE associated with a random (but present) page
   within the guest's kernel's address space
   (using a modified xc_translate_foreign_address)
3. Flips the present bit to zero
4. Calls HYPERVISOR_update_va_mapping otherdomain (from the kernel module)
5. Re-performs xc_translate_foreign_address on the same PTE
6. Verifies the the present bit was indeed changed to zero

Upon successful return from update_va_mapping (i.e. return code == zero),
my problem involves the last step mentioned above. The present bit still
remains unchanged (is still one), although update_va_mapping return true.

What might cause such a strange problem to occur?

-- 
/*
Michael R. Hines
http://www.cs.binghamton.edu/~mhines/
Live long and prosper...
*/


[-- Attachment #1.2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: update_va_mapping_otherdomain

[Keir Fraser]

On 8/9/07 02:29, "Michael R. Hines" <mhines@cs.binghamton.edu> wrote:

> Upon successful return from update_va_mapping (i.e. return code == zero),
> my problem involves the last step mentioned above. The present bit still
> remains unchanged (is still one), although update_va_mapping return true.
> 
> What might cause such a strange problem to occur?

That hypercall allows you to update a *local* va mapping to map *another*
domain's memory page. So you've mapped the guest's page somewhere random in
your own address space.

 -- Keir

Re: update_va_mapping_otherdomain

[Michael R. Hines]

[-- Attachment #1.1: Type: text/plain, Size: 953 bytes --]


ooooh, ok. I understand now.

Thank you very much for your response. =)

- Michael

Keir Fraser wrote:
> On 8/9/07 02:29, "Michael R. Hines" <mhines@cs.binghamton.edu> wrote:
> 
>> Upon successful return from update_va_mapping (i.e. return code == zero),
>> my problem involves the last step mentioned above. The present bit still
>> remains unchanged (is still one), although update_va_mapping return true.
>>
>> What might cause such a strange problem to occur?
> 
> That hypercall allows you to update a *local* va mapping to map *another*
> domain's memory page. So you've mapped the guest's page somewhere random in
> your own address space.
> 
>  -- Keir
> 
> 
> 
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel
> 

-- 
/*
Michael R. Hines
http://www.cs.binghamton.edu/~mhines/
Live long and prosper...
*/


[-- Attachment #1.2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: update_va_mapping_otherdomain

[Michael R. Hines]

[-- Attachment #1.1: Type: text/plain, Size: 1535 bytes --]


So, if I would like to use dom0 to modify a specific L1 PTE belonging
to another unprivileged domain, what's the best way to go about it?

Grant references? (on L1 page tables?)

Might I be able to get a quick pointer on the right sequence of hypercalls?

Much appreciated,
- Michael

Michael R. Hines wrote:
> ooooh, ok. I understand now.
> 
> Thank you very much for your response. =)
> 
> - Michael
> 
> Keir Fraser wrote:
>> On 8/9/07 02:29, "Michael R. Hines" <mhines@cs.binghamton.edu> wrote:
>>
>>> Upon successful return from update_va_mapping (i.e. return code == zero),
>>> my problem involves the last step mentioned above. The present bit still
>>> remains unchanged (is still one), although update_va_mapping return true.
>>>
>>> What might cause such a strange problem to occur?
>> That hypercall allows you to update a *local* va mapping to map *another*
>> domain's memory page. So you've mapped the guest's page somewhere random in
>> your own address space.
>>
>>  -- Keir
>>
>>
>>
>>
>> _______________________________________________
>> Xen-devel mailing list
>> Xen-devel@lists.xensource.com
>> http://lists.xensource.com/xen-devel
>>
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel

-- 
/*
Michael R. Hines
http://www.cs.binghamton.edu/~mhines/
Live long and prosper...
*/


[-- Attachment #1.2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: update_va_mapping_otherdomain

[Keir Fraser]

There is currently no way to modify active page tables of another domain.

 -- Keir


On 8/9/07 19:27, "Michael R. Hines" <mhines@cs.binghamton.edu> wrote:

> 
> So, if I would like to use dom0 to modify a specific L1 PTE belonging
> to another unprivileged domain, what's the best way to go about it?
> 
> Grant references? (on L1 page tables?)
> 
> Might I be able to get a quick pointer on the right sequence of hypercalls?
> 
> Much appreciated,
> - Michael
> 
> Michael R. Hines wrote:
>> ooooh, ok. I understand now.
>> 
>> Thank you very much for your response. =)
>> 
>> - Michael
>> 
>> Keir Fraser wrote:
>>> On 8/9/07 02:29, "Michael R. Hines" <mhines@cs.binghamton.edu> wrote:
>>> 
>>>> Upon successful return from update_va_mapping (i.e. return code == zero),
>>>> my problem involves the last step mentioned above. The present bit still
>>>> remains unchanged (is still one), although update_va_mapping return true.
>>>> 
>>>> What might cause such a strange problem to occur?
>>> That hypercall allows you to update a *local* va mapping to map *another*
>>> domain's memory page. So you've mapped the guest's page somewhere random in
>>> your own address space.
>>> 
>>>  -- Keir
>>> 
>>> 
>>> 
>>> 
>>> _______________________________________________
>>> Xen-devel mailing list
>>> Xen-devel@lists.xensource.com
>>> http://lists.xensource.com/xen-devel
>>> 
>> 
>> 
>> ------------------------------------------------------------------------
>> 
>> _______________________________________________
>> Xen-devel mailing list
>> Xen-devel@lists.xensource.com
>> http://lists.xensource.com/xen-devel

Re: update_va_mapping_otherdomain

[Michael R. Hines]

[-- Attachment #1.1: Type: text/plain, Size: 1862 bytes --]


Thank you.

- Michael

Keir Fraser wrote:
> There is currently no way to modify active page tables of another domain.
> 
>  -- Keir
> 
> 
> On 8/9/07 19:27, "Michael R. Hines" <mhines@cs.binghamton.edu> wrote:
> 
>> So, if I would like to use dom0 to modify a specific L1 PTE belonging
>> to another unprivileged domain, what's the best way to go about it?
>>
>> Grant references? (on L1 page tables?)
>>
>> Might I be able to get a quick pointer on the right sequence of hypercalls?
>>
>> Much appreciated,
>> - Michael
>>
>> Michael R. Hines wrote:
>>> ooooh, ok. I understand now.
>>>
>>> Thank you very much for your response. =)
>>>
>>> - Michael
>>>
>>> Keir Fraser wrote:
>>>> On 8/9/07 02:29, "Michael R. Hines" <mhines@cs.binghamton.edu> wrote:
>>>>
>>>>> Upon successful return from update_va_mapping (i.e. return code == zero),
>>>>> my problem involves the last step mentioned above. The present bit still
>>>>> remains unchanged (is still one), although update_va_mapping return true.
>>>>>
>>>>> What might cause such a strange problem to occur?
>>>> That hypercall allows you to update a *local* va mapping to map *another*
>>>> domain's memory page. So you've mapped the guest's page somewhere random in
>>>> your own address space.
>>>>
>>>>  -- Keir
>>>>
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> Xen-devel mailing list
>>>> Xen-devel@lists.xensource.com
>>>> http://lists.xensource.com/xen-devel
>>>>
>>>
>>> ------------------------------------------------------------------------
>>>
>>> _______________________________________________
>>> Xen-devel mailing list
>>> Xen-devel@lists.xensource.com
>>> http://lists.xensource.com/xen-devel
> 
> 
> 

-- 
/*
Michael R. Hines
http://www.cs.binghamton.edu/~mhines/
Live long and prosper...
*/


[-- Attachment #1.2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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