Re: [Xen-devel] [PATCH v4 04/14] argo: init, destroy and soft-reset, with enable command line opt

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 15.01.19 at 13:29, <roger.pau@xxxxxxxxxx> wrote:
> On Tue, Jan 15, 2019 at 01:27:36AM -0800, Christopher Clark wrote:
>>  long
>>  do_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1,
>>             XEN_GUEST_HANDLE_PARAM(void) arg2, unsigned long arg3,
>>             unsigned long arg4)
>>  {
>> -    return -ENOSYS;
>> +    long rc = -EFAULT;
>> +
>> +    argo_dprintk("->do_argo_op(%u,%p,%p,%lu,0x%lx)\n", cmd,
>> +                 (void *)arg1.p, (void *)arg2.p, arg3, arg4);
>> +
>> +    if ( unlikely(!opt_argo_enabled) )
>> +        return -EOPNOTSUPP;
> 
> I think this should return -ENOSYS, an hypervisor built with
> CONFIG_ARGO but without argo enabled on the command line shouldn't
> behave differently than an hypervisor build without CONFIG_ARGO.

We've been there before, and there appears to be disagreement.
I support the use of -EOPNOTSUPP here.

>> --- a/xen/include/xen/sched.h
>> +++ b/xen/include/xen/sched.h
>> @@ -490,6 +490,11 @@ struct domain
>>          unsigned int guest_request_enabled       : 1;
>>          unsigned int guest_request_sync          : 1;
>>      } monitor;
>> +
>> +#ifdef CONFIG_ARGO
>> +    /* Argo interdomain communication support */
>> +    struct argo_domain *argo;
> 
> I'm likely missing something, but argo_domain is declared in argo.c,
> don't you need a forward declaration here for this to build?

That would be needed in C++ (iirc), but not in C, where such
forward declarations are needed solely when the type is used
as a function parameter, as otherwise that parameter's type
ends up having scope local to the declared function.

Jan



_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel