Re: [PATCH v3 4/5] x86: Port the basic alternative mechanism from Linux to Xen

["Jan Beulich" <JBeulich@suse.com>]

>>> On 03.06.14 at 12:35, <feng.wu@intel.com> wrote:

> 
>> -----Original Message-----
>> From: Jan Beulich [mailto:JBeulich@suse.com]
>> Sent: Tuesday, June 03, 2014 6:25 PM
>> To: Wu, Feng
>> Cc: andrew.cooper3@citrix.com; ian.campbell@citrix.com;
>> stefano.stabellini@citrix.com; keir.xen@gmail.com; xen-devel@lists.xen.org;
>> boris.ostrovsky@oracle.com; tim@xen.org 
>> Subject: RE: [PATCH v3 4/5] x86: Port the basic alternative mechanism from
>> Linux to Xen
>> 
>> >>> On 03.06.14 at 12:13, <feng.wu@intel.com> wrote:
>> >> -----Original Message-----
>> >> From: Jan Beulich [mailto:JBeulich@suse.com]
>> >> Sent: Tuesday, June 03, 2014 6:00 PM
>> >> To: Wu, Feng
>> >> Cc: andrew.cooper3@citrix.com; ian.campbell@citrix.com;
>> >> stefano.stabellini@citrix.com; keir.xen@gmail.com;
>> xen-devel@lists.xen.org;
>> >> boris.ostrovsky@oracle.com; tim@xen.org 
>> >> Subject: Re: [PATCH v3 4/5] x86: Port the basic alternative mechanism from
>> >> Linux to Xen
>> >>
>> >> >>> On 30.05.14 at 10:56, <feng.wu@intel.com> wrote:
>> >> > +#ifdef K8_NOP1
>> >> > +static const unsigned char k8nops[] __initconst = {
>> >> > +    K8_NOP1,
>> >> > +    K8_NOP2,
>> >> > +    K8_NOP3,
>> >> > +    K8_NOP4,
>> >> > +    K8_NOP5,
>> >> > +    K8_NOP6,
>> >> > +    K8_NOP7,
>> >> > +    K8_NOP8
>> >> > +};
>> >> > +static const unsigned char * const k8_nops[ASM_NOP_MAX+1]
>> __initconst =
>> >> {
>> >> > +    NULL,
>> >> > +    k8nops,
>> >> > +    k8nops + 1,
>> >> > +    k8nops + 1 + 2,
>> >> > +    k8nops + 1 + 2 + 3,
>> >> > +    k8nops + 1 + 2 + 3 + 4,
>> >> > +    k8nops + 1 + 2 + 3 + 4 + 5,
>> >> > +    k8nops + 1 + 2 + 3 + 4 + 5 + 6,
>> >> > +    k8nops + 1 + 2 + 3 + 4 + 5 + 6 + 7
>> >> > +};
>> >> > +#endif
>> >> > +
>> >> > +#ifdef P6_NOP1
>> >> > +static const unsigned char p6nops[] __initconst = {
>> >> > +    P6_NOP1,
>> >> > +    P6_NOP2,
>> >> > +    P6_NOP3,
>> >> > +    P6_NOP4,
>> >> > +    P6_NOP5,
>> >> > +    P6_NOP6,
>> >> > +    P6_NOP7,
>> >> > +    P6_NOP8
>> >> > +};
>> >> > +static const unsigned char * const p6_nops[ASM_NOP_MAX+1]
>> __initconst =
>> >> {
>> >> > +    NULL,
>> >> > +    p6nops,
>> >> > +    p6nops + 1,
>> >> > +    p6nops + 1 + 2,
>> >> > +    p6nops + 1 + 2 + 3,
>> >> > +    p6nops + 1 + 2 + 3 + 4,
>> >> > +    p6nops + 1 + 2 + 3 + 4 + 5,
>> >> > +    p6nops + 1 + 2 + 3 + 4 + 5 + 6,
>> >> > +    p6nops + 1 + 2 + 3 + 4 + 5 + 6 + 7
>> >> > +};
>> >> > +#endif
>> >>
>> >> The uses of __initconst together with -fPIC cause build problems on
>> >> older gcc. I fixed this up in a temporary way, but put a work item on
>> >> my todo list to deal with this properly while at once converting
>> >> alternative.o to alternative.init.o (as the file consists of only init
>> >> code and data).
>> >
>> > Thanks a lot for pointing this out. Do I need to do something for this?
>> 
>> If you're up to it, you could of course save me the time to do what I
>> outlined above...
> 
> From the above you mentioned, seems we cannot use __initconst together with 
> -fPIC on
> some old gcc. Do you know what is the reason, and does the problem occur 
> with "__initdata ", "__init ", etc?

The problem is only with __initconst afaict, the reason being that with
-fPIC constant data with relocations doesn't get put into r/o sections
(irrespective of the "const" modifier) by older gcc versions.

Jan