Peter Maydell <peter.maydell@xxxxxxxxxx> writes:
> On 11 September 2012 07:10, Rusty Russell <rusty.russell@xxxxxxxxxx> wrote:
>> We use space #18 for floating point regs.
>
> Userspace ABI looks fine to me.
>
>> +static unsigned int num_vfp_regs(void)
>> +{
>> + /* FP regs, FPEXC, FPSCR, FPINST, FPINST2, MVFR0, MVFR1, FPSID. */
>> + return num_fp_regs() + 7;
>> +}
>> +
>> +static int copy_vfp_regids(u64 __user *uindices)
>> +{
>> + unsigned int i;
>> + const u64 u32reg = KVM_REG_ARM | KVM_REG_SIZE_U32 | KVM_REG_ARM_VFP;
>> + const u64 u64reg = KVM_REG_ARM | KVM_REG_SIZE_U64 | KVM_REG_ARM_VFP;
>> +
>> + for (i = 0; i < num_fp_regs(); i++) {
>> + if (put_user((u64reg | KVM_REG_ARM_VFP_ID_BASE_REG) + i,
>> + uindices))
>> + return -EFAULT;
>> + uindices++;
>> + }
>> +
>> + if (put_user(u32reg | KVM_REG_ARM_VFP_ID_FPEXC, uindices))
>> + return -EFAULT;
>> + uindices++, i++;
>> + if (put_user(u32reg | KVM_REG_ARM_VFP_ID_FPSCR, uindices))
>> + return -EFAULT;
>> + uindices++, i++;
>> + if (put_user(u32reg | KVM_REG_ARM_VFP_ID_FPINST, uindices))
>> + return -EFAULT;
>> + uindices++, i++;
>> + if (put_user(u32reg | KVM_REG_ARM_VFP_ID_FPINST2, uindices))
>> + return -EFAULT;
>> + uindices++, i++;
>> + if (put_user(u32reg | KVM_REG_ARM_VFP_ID_MVFR0, uindices))
>> + return -EFAULT;
>> + uindices++, i++;
>> + if (put_user(u32reg | KVM_REG_ARM_VFP_ID_MVFR1, uindices))
>> + return -EFAULT;
>> + uindices++, i++;
>> + if (put_user(u32reg | KVM_REG_ARM_VFP_ID_FPSID, uindices))
>> + return -EFAULT;
>> + uindices++, i++;
>> +
>> + return i;
>> +}
>
> If you had an array:
> static const int vfp_sysregs[] = { KVM_REG_ARM_VFP_ID_FPEXC,
> KVM_REG_ARM_VFP_ID_FPSCR, etc etc etc };
>
> then you could get rid of the hardcoded 7 in num_vfp_regs()
> and have a loop in copy_vfp_regids() rather than the repeated
> put_user() stanzas. That seems to me like it would be slightly
> cleaner-looking code.
Agreed. This was OK with 4 regs (as I started with), not so much now.
Thanks,
Rusty.
_______________________________________________
kvmarm mailing list
kvmarm@xxxxxxxxxxxxxxxxxxxxx
https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
