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

Re: [PATCH] libxc/PM: correct (not just) error handling in xc_get_cpufreq_para()


  • To: Anthony PERARD <anthony.perard@xxxxxxxxxx>
  • From: Jan Beulich <jbeulich@xxxxxxxx>
  • Date: Fri, 28 Mar 2025 12:06:49 +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: xen-devel@xxxxxxxxxxxxxxxxxxxx, Juergen Gross <jgross@xxxxxxxx>
  • Delivery-date: Fri, 28 Mar 2025 11:06:59 +0000
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>

On 28.03.2025 11:51, Anthony PERARD wrote:
> On Thu, Mar 27, 2025 at 02:32:04PM +0100, Jan Beulich wrote:
>> HWP work made the cleanup of the "available governors" array
>> conditional, neglecting the fact that the condition used may not be the
> 
> I don't know why the cleanup was made conditional, as long as the bounce
> buffer is declared with DECLARE_NAMED_HYPERCALL_BOUNCE(),
> xc_hypercall_bounce_post() can be called without issue (even if
> ..bounce_pre isn't used). Maybe it's all those "unlock_*" labels that is
> misleading, a single "out" label which does the cleanup
> unconditionally would be more than enough.

Oh, yet more cleanup to do there (independently of course).

>> condition that was used to allocate the buffer (as the structure field
>> is updated upon getting back EAGAIN). Throughout the function, use the
>> local variable being introduced to address that.
>>
>> --- a/tools/libs/ctrl/xc_pm.c
>> +++ b/tools/libs/ctrl/xc_pm.c
>> @@ -212,31 +212,32 @@ int xc_get_cpufreq_para(xc_interface *xc
>>      DECLARE_NAMED_HYPERCALL_BOUNCE(scaling_available_governors,
>>                       user_para->scaling_available_governors,
>>                       user_para->gov_num * CPUFREQ_NAME_LEN * sizeof(char), 
>> XC_HYPERCALL_BUFFER_BOUNCE_BOTH);
>> -
>> -    bool has_num = user_para->cpu_num &&
>> -                     user_para->freq_num &&
>> -                     user_para->gov_num;
>> +    unsigned int in_gov_num = user_para->gov_num;
>> +    bool has_num = user_para->cpu_num && user_para->freq_num;
>>  
>>      if ( has_num )
> 
> I think there's an issue here, this condition used to be true if
> `gov_num` was not 0, even if `cpu_num` and `freq_num` was 0. That's not
> the case anymore. Shouldn't `has_num` use also the value from
> `gov_num`? Do we actually need `has_num`?

Raising this question is where all of this started:
https://lists.xen.org/archives/html/xen-devel/2025-03/msg01870.html.
IOW with Penny's change I think the need for has_num will disappear, the
latest. At this point, requesting the governors being optional, only
->gov_num shouldn't affect has_num. Once requesting frequencies becomes
optional too, has_num would become a mere alias of ->cpu_num.

Jan



 


Rackspace

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