Re: RT Xen on ARM - R-Car series

[Julien Grall <julien.grall@arm.com>]

Hi,

On 2/4/19 12:53 PM, Roger Pau Monné wrote:
> On Fri, Feb 01, 2019 at 05:40:14PM +0000, Julien Grall wrote:
>> Hi,
>>
>> On 01/02/2019 16:53, Roger Pau Monné wrote:
>>> On Thu, Jan 31, 2019 at 11:14:37PM +0000, Julien Grall wrote:
>>>> On 1/31/19 9:56 PM, Stefano Stabellini wrote:
>>>>> On Thu, 31 Jan 2019, Julien Grall wrote:
>>>>>> On 31/01/2019 12:00, Andrii Anisov wrote:
>>>>>>> On 31.01.19 13:37, Julien Grall wrote:
>>> So, I've got a hacky patch to 'fix' this on x86, by taking a reference
>>> to the mfn behind the virtual address provided when setting up the
>>> hypercall and mapping it in Xen.
>>
>> That was the idea I had in mind :). Hopefully, no guest is modifying the
>> mapping (i.e the virtual address point to a different physical address)
>> afterwards.
>>
>>> This however doesn't work on ARM due
>>> to the lack of paging_gva_to_gfn. I guess there's something similar to
>>> translate a guest virtual address into a gfn or a mfn?
>>
>> get_page_from_gva should to the job for you.
>>> +int map_runstate_area(struct vcpu *v,
>>> +                      struct vcpu_register_runstate_memory_area *area)
>>> +{
>>> +    unsigned long offset;
>>> +    unsigned int i;
>>> +    struct domain *d = v->domain;
>>> +    size_t size =
>>> +#ifdef CONFIG_COMPAT
>>> +        has_32bit_shinfo((v)->domain) ? sizeof(*v->compat_runstate_guest) :
>>> +#endif
>>> +                                        sizeof(*v->runstate_guest);
>>> +
>>> +    if ( v->runstate_guest )
>>> +    {
>>> +        ASSERT_UNREACHABLE();
>>> +        return -EBUSY;
>>> +    }
>>> +
>>> +    offset = area->addr.p & ~PAGE_MASK;
>>> +
>>> +    for ( i = 0; i < ARRAY_SIZE(v->runstate_mfn); i++ )
>>> +    {
>>> +        p2m_type_t t;
>>> +        uint32_t pfec = PFEC_page_present;
>>> +        gfn_t gfn = _gfn(paging_gva_to_gfn(v, area->addr.p, &pfec));
>>> +        struct page_info *pg;
>>> +
>>> +        if ( gfn_eq(gfn, INVALID_GFN) )
>>> +            return -EFAULT;
>>> +
>>> +        v->runstate_mfn[i] = get_gfn(d, gfn_x(gfn), &t);
>>
>> get_gfn would need to be implemented on Arm.
>>
>>> +        if ( t != p2m_ram_rw || mfn_eq(v->runstate_mfn[i], INVALID_MFN) )
>>> +            return -EFAULT;
>>> +
>>> +        pg = mfn_to_page(v->runstate_mfn[i]);
>>> +        if ( !pg || !get_page_type(pg, PGT_writable_page) )
>>> +        {
>>> +            put_gfn(d, gfn_x(gfn));
>>> +            return -EFAULT;
>>> +        }
>>> +
>>> +        put_gfn(d, gfn_x(gfn));
>>> +
>>> +        if ( PFN_DOWN(gfn_x(gfn) + size) == PFN_DOWN(gfn_x(gfn)) )
>>
>> This looks wrong, you seem to mix address and frame. I think you might want:
>>
>> if ( gfn_eq(gfn_add(gfn, PFN_DOWN(size)), gfn) )
> 
> Thanks!
> 
> Here is an updated version which seems to build on ARM. I don't have
> an easy way to test this, could you give it a spin?

Thank you for writing a patch. I am back in France this week for family 
reason and will not have time properly give a spin this week. Stefano, 
Andrii, can you test it?

> 
> I don't like adding CONFIG_X86/ARM in common code, so it might be
> worth to either try to factor this out into arch specific code, or
> even better, provide common functions to translate a guest virtual
> address into a gfn, mfn or page.

I would prefer GVA to MFN/page because we already have a fairly 
complicate helper handling all the cases (e.g memaccess, 
break-before-make, ...) for translation and taking a reference on the 
page. Although we could potentially split the function in two if we want 
to cater any other translation.

One comment below.

> +#elif defined(CONFIG_ARM)
> +        pg = get_page_from_gva(v, area->addr.p, GV2M_WRITE);
> +        if ( !pg || !get_page_type(pg, PGT_writable_page) )

One reference is already taken by get_page_from_gva. So you should not 
need to take another here.

Cheers,

-- 
Julien Grall

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xenproject.org
https://lists.xenproject.org/mailman/listinfo/xen-devel