Re: [Qemu-devel] testing qemu svn r6082 on FreeBSD - virtio-net, hpet, vmmouse/vga, got bsd-user to build, and an updated version of the FreeBSD -clock dynticks patch

[Juergen Lock]

In article <address@hidden> you write:
>On 12/19/08, Juergen Lock <address@hidden> wrote:
>>  +#elif defined(__FreeBSD__)
>>  +# include <sys/ucontext.h>
>>  +
>>  +# define RIP_sig(context)  (*((unsigned 
>> long*)&(context)->uc_mcontext.mc_rip))
>>  +# define TRAP_sig(context)    ((context)->uc_mcontext.mc_trapno)
>>  +# define ERROR_sig(context)   ((context)->uc_mcontext.mc_err)
>>   #else
>>   #define QEMU_UC_MCONTEXT_GREGS(uc, reg)        
>> (uc)->uc_mcontext.gregs[(reg)]
>>   #define QEMU_UC_MACHINE_PC(uc)         QEMU_UC_MCONTEXT_GREGS(uc, REG_RIP)
>
>>  +#ifdef __FreeBSD__
>>  +    pc = RIP_sig(uc);
>>  +    return handle_cpu_signal(pc, (unsigned long)info->si_addr,
>>  +                             TRAP_sig(uc) == 0xe ?
>>  +                             (ERROR_sig(uc) >> 1) & 1 : 0,
>>  +                             &uc->uc_sigmask, puc);
>>  +#else
>>      pc = QEMU_UC_MACHINE_PC(uc);
>>      return handle_cpu_signal(pc, (unsigned long)info->si_addr,
>>                               QEMU_UC_MCONTEXT_GREGS(uc, REG_TRAPNO) == 0xe ?
>>                               (QEMU_UC_MCONTEXT_GREGS(uc, REG_ERR) >> 1) & 1 
>> : 0,
>>                               &uc->uc_sigmask, puc);
>>  +#endif
>
>The idea here was that all OS define macros with same names so that
>the code below does not get any more complex. Maybe the GREGS macro
>was too generic, and should be replaced with one that only returns the
>trap and error values.
>
Yeah I was too lazy to figure out the preprocessor magic needed
to get the GREGS way working so I simply reused the macros from the
i386 case. :)

>>   And finally the updated dynticks patch:
>
>This looks OK, please submit separately.

 OK will do.

 Thanx,
        Juergen