Re: [Xen-devel] [PATCH RFC] xen: arm: Log a warning message when a deprecated hypercall is used

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 20.01.15 at 13:05, <Ian.Campbell@xxxxxxxxxx> wrote:
> On Tue, 2015-01-20 at 11:19 +0000, Jan Beulich wrote:
>> >>> On 20.01.15 at 12:11, <Ian.Campbell@xxxxxxxxxx> wrote:
>> > On Tue, 2015-01-20 at 11:05 +0000, Jan Beulich wrote:
>> >> >>> On 20.01.15 at 11:52, <ian.campbell@xxxxxxxxxx> wrote:
>> >> > +static register_t do_deprecated_hypercall(void)
>> >> > +{
>> >> > +    struct cpu_user_regs *regs = guest_cpu_user_regs();
>> >> > +    const register_t op =
>> >> > +#ifdef CONFIG_ARM_64
>> >> > +        !is_32bit_domain(current->domain) ?
>> >> > +            regs->x16
>> >> > +        :
>> >> > +#endif
>> >> > +            regs->r12;
>> >> > +
>> >> > +    gdprintk(XENLOG_DEBUG, "%pv: deprecated hypercall %ld\n",
>> >> > +             current, (unsigned long)op);
>> >> 
>> >> If this was x86 code, I'd complain about the cast...
>> > 
>> > The correct format code is PRIregister (since register_t can be 32- or
>> > 64-bit for arm32 vs arm64 respectively), but it is in hex and xen.h
>> > lists __HYPERVISOR_* in decimal so that's what I wanted to print for
>> > each of manually looking.
>> 
>> Which makes sense. I'd rather question whether PRIregister is
>> badly defined then, in that it not just specifies an eventual size
>> prefix. I.e. have it be just like C99's PRIx64 and alike, commonly
>> implemented by a (library internal) abstraction just holding the
>> necessary size prefix.
> 
> Yes, it probably is, it's just that 99% of the time hex is what we want
> so we haven't hit the activation energy to go and fix them.
> 
> Your suggestion is PRIxREGISTER and PRIdREGISTER etc, right?

PRIxREG etc perhaps.

> There are several other non-ARM specific ones, PRIpaddr, PRI_xen_pfn and
> PRI_xen_ulong etc (which may well have been added by me too during the
> ARM port, so I won't claim them as precedent). Should we be changing
> them all? I suppose paddr and pfn are less likely to be desired in
> decimal.

Right - out of these only the ulong one may want to get relaxed.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel