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

Re: [PATCH] misra: fix violations in macros GVA_INFO, TRACE_TIME


  • To: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>, Dmytro Prokopchuk1 <dmytro_prokopchuk1@xxxxxxxx>
  • From: Jan Beulich <jbeulich@xxxxxxxx>
  • Date: Thu, 31 Jul 2025 17:58:12 +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: Stefano Stabellini <sstabellini@xxxxxxxxxx>, Julien Grall <julien@xxxxxxx>, Bertrand Marquis <bertrand.marquis@xxxxxxx>, Michal Orzel <michal.orzel@xxxxxxx>, Volodymyr Babchuk <Volodymyr_Babchuk@xxxxxxxx>, Dario Faggioli <dfaggioli@xxxxxxxx>, Juergen Gross <jgross@xxxxxxxx>, George Dunlap <gwd@xxxxxxxxxxxxxx>, "xen-devel@xxxxxxxxxxxxxxxxxxxx" <xen-devel@xxxxxxxxxxxxxxxxxxxx>
  • Delivery-date: Thu, 31 Jul 2025 15:58:31 +0000
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>

On 31.07.2025 17:37, Andrew Cooper wrote:
> On 31/07/2025 4:16 pm, Dmytro Prokopchuk1 wrote:
>> MISRA Rule 13.1: Initializer lists shall not contain persistent side
>> effects.
>>
>> The violations occur because both the `GVA_INFO` and `TRACE_TIME` macro
>> expansions include expressions with persistent side effects introduced
>> via inline assembly.
>>
>> In the case of `GVA_INFO`, the issue stems from the initializer list
>> containing a direct call to `current`, which evaluates to
>> `this_cpu(curr_vcpu)` and involves persistent side effects via the
>> `asm` statement. To resolve this, the side-effect-producing expression
>> is computed in a separate statement prior to the macro initialization:
>>
>>     struct vcpu *current_vcpu = current;
>>
>> The computed value is passed into the `GVA_INFO(current_vcpu)` macro,
>> ensuring that the initializer is clean and free of such side effects.
>>
>> Similarly, the `TRACE_TIME` macro violates this rule when accessing
>> expressions like `current->vcpu_id` and `current->domain->domain_id`,
>> which also depend on `current` and inline assembly. To fix this, the
>> value of `current` is assigned to a temporary variable:
>>
>>     struct vcpu *v = current;
>>
>> This temporary variable is then used to access `domain_id` and `vcpu_id`.
>> This ensures that the arguments passed to the `TRACE_TIME` macro are
>> simple expressions free of persistent side effects.
>>
>> Signed-off-by: Dmytro Prokopchuk <dmytro_prokopchuk1@xxxxxxxx>
> 
> The macro `current` specifically does not (and must not) have side
> effects.  It is expected to behave like a plain `struct vcpu *current;`
> variable, and what Eclair is noticing is the thread-local machinery
> under this_cpu() (or in x86's case, get_current()).
> 
> In ARM's case, it's literally reading the hardware thread pointer
> register.  Can anything be done to tell Eclair that `this_cpu()`
> specifically does not have side effects?
> 
> The only reason that GVA_INFO() and TRACE_TIME() are picked out is
> because they both contain embedded structure initialisation, and this is
> is actually an example where trying to comply with MISRA interferes with
> what is otherwise a standard pattern in Xen.

Irrespective of what you say, some of the changes here were eliminating
multiple adjacent uses of current, which - iirc - often the compiler
can't fold via CSE.

Jan



 


Rackspace

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