[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH] x86/hvm: Add APIC IDs to the per-vLAPIC save area


  • To: Roger Pau Monné <roger.pau@xxxxxxxxxx>, Alejandro Vallejo <alejandro.vallejo@xxxxxxxxx>
  • From: Jan Beulich <jbeulich@xxxxxxxx>
  • Date: Thu, 27 Feb 2025 08:29:24 +0100
  • Autocrypt: addr=jbeulich@xxxxxxxx; keydata= xsDiBFk3nEQRBADAEaSw6zC/EJkiwGPXbWtPxl2xCdSoeepS07jW8UgcHNurfHvUzogEq5xk hu507c3BarVjyWCJOylMNR98Yd8VqD9UfmX0Hb8/BrA+Hl6/DB/eqGptrf4BSRwcZQM32aZK 7Pj2XbGWIUrZrd70x1eAP9QE3P79Y2oLrsCgbZJfEwCgvz9JjGmQqQkRiTVzlZVCJYcyGGsD /0tbFCzD2h20ahe8rC1gbb3K3qk+LpBtvjBu1RY9drYk0NymiGbJWZgab6t1jM7sk2vuf0Py O9Hf9XBmK0uE9IgMaiCpc32XV9oASz6UJebwkX+zF2jG5I1BfnO9g7KlotcA/v5ClMjgo6Gl MDY4HxoSRu3i1cqqSDtVlt+AOVBJBACrZcnHAUSuCXBPy0jOlBhxPqRWv6ND4c9PH1xjQ3NP nxJuMBS8rnNg22uyfAgmBKNLpLgAGVRMZGaGoJObGf72s6TeIqKJo/LtggAS9qAUiuKVnygo 3wjfkS9A3DRO+SpU7JqWdsveeIQyeyEJ/8PTowmSQLakF+3fote9ybzd880fSmFuIEJldWxp Y2ggPGpiZXVsaWNoQHN1c2UuY29tPsJgBBMRAgAgBQJZN5xEAhsDBgsJCAcDAgQVAggDBBYC AwECHgECF4AACgkQoDSui/t3IH4J+wCfQ5jHdEjCRHj23O/5ttg9r9OIruwAn3103WUITZee e7Sbg12UgcQ5lv7SzsFNBFk3nEQQCACCuTjCjFOUdi5Nm244F+78kLghRcin/awv+IrTcIWF hUpSs1Y91iQQ7KItirz5uwCPlwejSJDQJLIS+QtJHaXDXeV6NI0Uef1hP20+y8qydDiVkv6l IreXjTb7DvksRgJNvCkWtYnlS3mYvQ9NzS9PhyALWbXnH6sIJd2O9lKS1Mrfq+y0IXCP10eS FFGg+Av3IQeFatkJAyju0PPthyTqxSI4lZYuJVPknzgaeuJv/2NccrPvmeDg6Coe7ZIeQ8Yj t0ARxu2xytAkkLCel1Lz1WLmwLstV30g80nkgZf/wr+/BXJW/oIvRlonUkxv+IbBM3dX2OV8 AmRv1ySWPTP7AAMFB/9PQK/VtlNUJvg8GXj9ootzrteGfVZVVT4XBJkfwBcpC/XcPzldjv+3 HYudvpdNK3lLujXeA5fLOH+Z/G9WBc5pFVSMocI71I8bT8lIAzreg0WvkWg5V2WZsUMlnDL9 mpwIGFhlbM3gfDMs7MPMu8YQRFVdUvtSpaAs8OFfGQ0ia3LGZcjA6Ik2+xcqscEJzNH+qh8V m5jjp28yZgaqTaRbg3M/+MTbMpicpZuqF4rnB0AQD12/3BNWDR6bmh+EkYSMcEIpQmBM51qM EKYTQGybRCjpnKHGOxG0rfFY1085mBDZCH5Kx0cl0HVJuQKC+dV2ZY5AqjcKwAxpE75MLFkr wkkEGBECAAkFAlk3nEQCGwwACgkQoDSui/t3IH7nnwCfcJWUDUFKdCsBH/E5d+0ZnMQi+G0A nAuWpQkjM1ASeQwSHEeAWPgskBQL
  • Cc: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>, xen-devel@xxxxxxxxxxxxxxxxxxxx
  • Delivery-date: Thu, 27 Feb 2025 07:29:31 +0000
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>

On 26.02.2025 18:33, Roger Pau Monné wrote:
> On Wed, Feb 26, 2025 at 02:11:23PM +0100, Jan Beulich wrote:
>> On 18.02.2025 15:22, Alejandro Vallejo wrote:
>>> Today, Xen hardcodes apic_id = vcpu_id * 2, but this is unwise and
>>> interferes with providing accurate topology information to the guest.
>>>
>>> Introduce a new x2apic_id field into hvm_hw_lapic.  This is immutable
>>> state from the guest's point of view, but it will allow the toolstack to
>>> eventually configure the value, and for the value to move on migrate.
>>>
>>> For backwards compatibility, the patch rebuilds the old-style APIC IDs
>>> from migration streams lacking them when they aren't present.
>>
>> Nit: "when they aren't present" looks to duplicate "lacking them"?
>>
>>> Signed-off-by: Alejandro Vallejo <alejandro.vallejo@xxxxxxxxx>
>>> ---
>>> I've split this one from the rest of the topology series as it's independent
>>> and entangled with another patch from Andrew.
>>
>> Albeit I think meanwhile we've settled that the entangling isn't quite as
>> problematic.
>>
>>> @@ -1621,6 +1624,14 @@ static int cf_check lapic_load_hidden(struct domain 
>>> *d, hvm_domain_context_t *h)
>>>          return -EINVAL;
>>>      }
>>>  
>>> +    /*
>>> +     * Xen 4.20 and earlier had no x2APIC ID in the migration stream and
>>> +     * hard-coded "vcpu_id * 2". Default back to this if we have a
>>> +     * zero-extended record.
>>> +     */
>>> +    if ( h->size <= offsetof(struct hvm_hw_lapic, x2apic_id) )
>>> +        s->hw.x2apic_id = v->vcpu_id * 2;
>>
>> While we better wouldn't get to see such input, it is in principle possible
>> to have an input stream with, say, half the field. Imo the condition ought
>> to be such that we'd make the adjustment when less than the full field is
>> available.
> 
> I would add an additional check to ensure _rsvd0 remains 0, to avoid
> further additions from attempting to reuse that padding space.
> 
> if ( s->hw._rsvd0 )
>     return -EINVAL;

I agree we want such a check; I actually should have pointed that out, too.
I don't, however, see why the field couldn't be re-used going forward (under
the right conditions, of course).

> In fact I would be tempted to overwrite the ID if the stream size
> doesn't match the expected one, ie:
> 
> if ( h->size < (offsetof(struct hvm_hw_lapic, _rsvd0) +
>                 sizeof(s->hw._rsvd0)) )
>     s->hw.x2apic_id = v->vcpu_id * 2;

Hmm, yes, perhaps better to be yet more safe here.

Jan



 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.