On Mon, 8 Mar 2021 at 20:14, Aaron Lindsay <aa...@os.amperecomputing.com> wrote:
>
> This allows plugins to query for full virtual-to-physical address
> translation for a given `qemu_plugin_hwaddr` and stops exposing the
> offset within the device itself. As this change breaks the API,
> QEMU_PLUGIN_VERSION is incremented.
>
> Signed-off-by: Aaron Lindsay <aa...@os.amperecomputing.com>
> ---
> diff --git a/include/qemu/qemu-plugin.h b/include/qemu/qemu-plugin.h
> index c66507fe8f..2252ecf2f0 100644
> --- a/include/qemu/qemu-plugin.h
> +++ b/include/qemu/qemu-plugin.h
> @@ -47,7 +47,7 @@ typedef uint64_t qemu_plugin_id_t;
>
> extern QEMU_PLUGIN_EXPORT int qemu_plugin_version;
>
> -#define QEMU_PLUGIN_VERSION 0
> +#define QEMU_PLUGIN_VERSION 1
>
> typedef struct {
> /* string describing architecture */
> @@ -328,7 +328,7 @@ struct qemu_plugin_hwaddr
> *qemu_plugin_get_hwaddr(qemu_plugin_meminfo_t info,
> * offset will be into the appropriate block of RAM.
> */
> bool qemu_plugin_hwaddr_is_io(const struct qemu_plugin_hwaddr *haddr);
> -uint64_t qemu_plugin_hwaddr_device_offset(const struct qemu_plugin_hwaddr
> *haddr);
> +uint64_t qemu_plugin_hwaddr_phys_addr(const struct qemu_plugin_hwaddr
> *haddr);
This should have a documentation comment, since this is the public-facing API.
Also, physical addresses aren't uniquely identifying, they're only valid
in the context of a particular address space (think TrustZone, for instance),
so the plugin-facing API should probably have some concept of how it
distinguishes "this is an access for Secure 0x4000" from "this is an
access for Non-Secure 0x4000".
thanks
-- PMM
