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

Re: [PATCH for-4.19] x86/altcall: fix clang code-gen when using altcall in loop constructs


  • To: Roger Pau Monné <roger.pau@xxxxxxxxxx>
  • From: Jan Beulich <jbeulich@xxxxxxxx>
  • Date: Wed, 24 Jul 2024 10:41:43 +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: Alejandro Vallejo <alejandro.vallejo@xxxxxxxxx>, xen-devel@xxxxxxxxxxxxxxxxxxxx, Andrew Cooper <andrew.cooper3@xxxxxxxxxx>, Oleksii Kurochko <oleksii.kurochko@xxxxxxxxx>
  • Delivery-date: Wed, 24 Jul 2024 08:41:53 +0000
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>

On 24.07.2024 10:28, Roger Pau Monné wrote:
> The only way I've found to cope with this is to use something like:
> 
> #define ALT_CALL_ARG(arg, n)                                            \
>     union {                                                             \
>         typeof(arg) e[(sizeof(long) + sizeof(arg) - 1) / sizeof(arg)];  \
>         unsigned long r;                                                \
>     } a ## n ## __  = {                                                 \
>         .e[0] = ({ BUILD_BUG_ON(sizeof(arg) > sizeof(void *)); (arg); })\
>     };                                                                  \
>     register unsigned long a ## n ## _ asm ( ALT_CALL_arg ## n ) =      \
>         a ## n ## __.r
> 
> An oversized array that ensures all the space of the long is covered
> by the array, but then we need an extra variable, as we would
> otherwise require modifying ALT_CALL{0..6}_OUT to use aX_.r instead of
> aX_.

I don't think we need to over-size the array. It just gets yet a little
more complex then:

#define ALT_CALL_ARG(arg, n)                                            \
    register union {                                                    \
        struct {                                                        \
            typeof(arg) e;                                              \
            char pad[sizeof(long) - sizeof(arg)];                       \
        } s;                                                            \
        unsigned long r;                                                \
    } a ## n ## _ asm ( ALT_CALL_arg ## n ) = {                         \
        .s.e = ({ BUILD_BUG_ON(sizeof(arg) > sizeof(void *)); (arg); }) \
    }

Jan



 


Rackspace

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