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

Re: [PATCH] x86/boot: Further simplify CR4 handling in dom0_construct_pv()


  • To: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>
  • From: Jan Beulich <jbeulich@xxxxxxxx>
  • Date: Mon, 7 Oct 2024 09:21:52 +0200
  • 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: Roger Pau Monné <roger.pau@xxxxxxxxxx>, Xen-devel <xen-devel@xxxxxxxxxxxxxxxxxxxx>
  • Delivery-date: Mon, 07 Oct 2024 07:21:59 +0000
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>

On 04.10.2024 20:49, Andrew Cooper wrote:
> On 04/10/2024 7:52 am, Jan Beulich wrote:
>> On 03.10.2024 01:20, Andrew Cooper wrote:
>>> The logic would be more robust disabling SMAP based on its precense in CR4,
>>> rather than SMAP's accociation with a synthetic feature.
>> It's hard to tell what's more robust without knowing what future changes
>> there might be. In particular ...
>>
>>> @@ -1064,19 +1065,19 @@ int __init dom0_construct_pv(struct domain *d,
>>>       * prevents us needing to write construct_dom0() in terms of
>>>       * copy_{to,from}_user().
>>>       */
>>> -    if ( boot_cpu_has(X86_FEATURE_XEN_SMAP) )
>>> +    if ( cr4 & X86_CR4_SMAP )
>> ... with this adjustment ...
>>
>>>      {
>>>          if ( IS_ENABLED(CONFIG_PV32) )
>>>              cr4_pv32_mask &= ~X86_CR4_SMAP;
>> ... this update of a global no longer occurs. Playing games with CR4
>> elsewhere might run into issues with this lack of updating.
> 
> We don't know the future, but I'm confused by your reasoning here. 
> Right now there's an expectation/assumption that FEAT_XEN_SMAP == CR4.SMAP.
> 
> In fact, the logic in staging right now is wonky if FEAT_XEN_SMAP=1 but
> CR4.SMAP=1.  In this case, we'll do nothing on the way in, and then
> activate SMAP on the way out.

I assume you meant "but CR4.SMAP=0". In that case yes, the logic here would
(kind of as a side effect) correct the wrong combination of state.

> construct_dom0() will definitely crash if SMAP is active.  So looking at
> CR4 is strictly better than accidentally falling into a FEAT_XEN_SMAP=0
> but CR4.SMAP=1 case.

It's better when taking one possible perspective, yes. Otoh CR4.SMAP=1 when
FEAT_XEN_SMAP=0 is a bug, and hence deserves being noticed (if nothing
else then by Xen crashing).

> Needing to play with the global cr4_pv32_mask is a consequence of
> choosing to disabling SMAP, rather than using STAC and/or rewriting
> using copy_*_user().  If you want to avoid playing with cr4_pv32_mask,
> we'll need to revisit this decision.
> 
> While the APs are active/working at this point in boot, they're not
> running guests (32bit PV or otherwise), so alterations to cr4_pv32_mask
> don't really matter.

I didn't really think of APs, but of the BSP itself.

Jan



 


Rackspace

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