On 20 August 2012 08:52, Alexander Graf <agraf@xxxxxxx> wrote: > Am 20.08.2012 um 06:57 schrieb Rusty Russell <rusty.russell@xxxxxxxxxx>: >> To be precise, you'd prefer: >> >> #define KVM_ARM_MSR_COPROC_MASK_START 16 >> #define KVM_ARM_MSR_COPROC_MASK_LEN 16 >> >> vs: >> #define KVM_ARM_MSR_COPROC_MASK 0xFFFF0000 >> >> What a huge readability lose. I can't describe how distasteful it is to >> do this because QEMU would have to write a macro :( > > The usual way this is done in the kernel is > > #define xxx_SHIFT 16 > #define xxx_MASK (0xffff << xxx_SHIFT) > > Right? Would that work for you guys? It'd certainly be easier to read. I now realise that the qemu code needs its macro to encode an index from a tuple independently of whether KVM is enabled, so it couldn't really use the kvm header macros anyway. (we might be able to throw in an assertion that they match up somewhere I guess). So you should probably use whatever standard kernel style is for telling userspace about aa set of bitmasks. -- PMM _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
