On Wed, Nov 26, 2014 at 05:46:05PM +0100, Paolo Bonzini wrote:

On 26/11/2014 15:58, Alex Benn?e wrote:

quoted

Andrew Jones [off-list ref] writes:

quoted

On Tue, Nov 25, 2014 at 04:10:00PM +0000, Alex Benn?e wrote:

quoted

This commit defines the API headers for guest debugging. There are two
architecture specific debug structures:

<snip>

quoted

quoted

+/* Architecture related debug defines - upper 16 bits of
+ * kvm_guest_debug->control
+ */
+#define KVM_GUESTDBG_USE_SW_BP_SHIFT	16
+#define KVM_GUESTDBG_USE_SW_BP		(1 << KVM_GUESTDBG_USE_SW_BP_SHIFT)
+#define KVM_GUESTDBG_USE_HW_BP_SHIFT	17
+#define KVM_GUESTDBG_USE_HW_BP		(1 << KVM_GUESTDBG_USE_HW_BP_SHIFT)
+

I see this are defined in arch/x86/include/uapi/asm/kvm.h,
so you needed to reproduce them here, but shouldn't they
be promoted to include/uapi/linux/kvm.h instead?

Well if we move them to common uapi we either restrict the $ARCH
specific options that don't have SW/HW BKPTS (would be weird but...) or
make them generic in the lower 16 bits (breaks API).

But in principle I have no objection if other don't.

I think it's a matter of personal taste.  "Architecture-specific" means
"not all architectures may support it", but it's certainly a good idea
to reuse the same value if multiple architectures do support a #define.

What you did is fine, another possibility is to do

    #define __KVM_GUESTDBG_USE_SW_BP   (1 << 16)

in include/uapi/linux/kvm.h, and

    #define KVM_GUESTDBG_USE_SW_BP __KVM_GUESTDBG_USE_SW_BP

in the arch-specific file.  Andrew, is this closer to what you intended?

I just reread Documentation/virtual/kvm/api.txt a bit more closely and
see that these fall in the "architecture specific control" region of the
field. So forget what I said. But your suggestion of __KVM_GUESTDBG_USE_SW_BP
looks like a good idea to me.

drew