Re: [Qemu-devel] [PATCH 03/12] Switch build system to accompanied kernel headers

[Jan Kiszka]

On 2011-06-14 13:28, Alexander Graf wrote:
> 
> On 14.06.2011, at 13:21, Jan Kiszka wrote:
> 
>> On 2011-06-14 13:11, Alexander Graf wrote:
>>>
>>> On 08.06.2011, at 16:10, Jan Kiszka wrote:
>>>
>>> This helps reducing our build-time checks for feature support in the
>>> available Linux kernel headers. And it helps users that do not have
>>> sufficiently recent headers installed on their build machine.
>>>
>>> Consequently, the patch removes and build-time checks for kvm and vhost
>>> in configure, the --kerneldir switch, and KVM_CFLAGS. Kernel headers are
>>> supposed to be provided by QEMU only.
>>>
>>> s390 needs some extra love as it carries redefinitions from kernel
>>> headers.
>>>
>>> Yes. I was wondering if we should unconditionally include the kernel 
>>> headers there. The problem I'm seeing there is that I don't know if that 
>>> would work fine on non-Linux hosts, as that code definitely gets compiled 
>>> there, while KVM code is not.
>>>
>>>
>>>
>>> [...]
>>>
>>> diff --git a/target-s390x/cpu.h b/target-s390x/cpu.h
>>> index 4e5c391..b5e587f 100644
>>> --- a/target-s390x/cpu.h
>>> +++ b/target-s390x/cpu.h
>>> @@ -313,16 +313,6 @@ CPUState *s390_cpu_addr2state(uint16_t cpu_addr);
>>> /* from s390-virtio-bus */
>>> extern const target_phys_addr_t virtio_size;
>>>
>>> -#ifndef KVM_S390_SIGP_STOP
>>> -#define KVM_S390_SIGP_STOP              0
>>> -#define KVM_S390_PROGRAM_INT            0
>>> -#define KVM_S390_SIGP_SET_PREFIX        0
>>> -#define KVM_S390_RESTART                0
>>> -#define KVM_S390_INT_VIRTIO             0
>>> -#define KVM_S390_INT_SERVICE            0
>>> -#define KVM_S390_INT_EMERGENCY          0
>>> -#endif
>>> -
>>> #endif
>>> void cpu_lock(void);
>>> void cpu_unlock(void);
>>> diff --git a/target-s390x/op_helper.c b/target-s390x/op_helper.c
>>> index db03a79..9429698 100644
>>> --- a/target-s390x/op_helper.c
>>> +++ b/target-s390x/op_helper.c
>>> @@ -23,6 +23,7 @@
>>> #include "helpers.h"
>>> #include <string.h>
>>> #include "kvm.h"
>>> +#include <linux/kvm.h>
>>>
>>> Have you tried to compile this on non-Linux?
>>
>> Sorry, I don't have "non-Linux" around. :) Do you expect build problems
>> with that header? Why?
> 
> Do you have a git tree handy? I can give a Mac build a try then.

Try git://git.kiszka.org/qemu-kvm.git queues/kvm-upstream

> 
>> BTW, if you depend on KVM_* constants for non-KVM builds, that looks a
>> bit fishy to me. Why is that code built at all in that setup?
> 
> The reason for that is that I didn't want to clutter the emulation code with 
> #ifdefs:
> 
> static void program_interrupt(CPUState *env, uint32_t code, int ilc)
> {
>     qemu_log("program interrupt at %#" PRIx64 "\n", env->psw.addr);
> 
>     if (kvm_enabled()) {
>         kvm_s390_interrupt(env, KVM_S390_PROGRAM_INT, code);
>     } else {
>         env->int_pgm_code = code;
>         env->int_pgm_ilc = ilc;
>         env->exception_index = EXCP_PGM;
>         cpu_loop_exit();
>     }
> }
> 
> This breaks compilation when KVM_S390_PROGRAM_INT is not defined. I'm very 
> open to suggestions on how to improve this though :).

Callbacks? See cpu_interrupt_handler e.g.

Jan

-- 
Siemens AG, Corporate Technology, CT T DE IT 1
Corporate Competence Center Embedded Linux