Re: [Qemu-devel] [PATCH 04/15] nvdimm: support nvdimm label

[Stefan Hajnoczi]

On Thu, Mar 17, 2016 at 04:32:50PM +0800, Xiao Guangrong wrote:
> +static void nvdimm_init(Object *obj)
> +{
> +    object_property_add_bool(obj, "reserve-label", nvdimm_get_reserve_label,
> +                             nvdimm_set_reserve_label, NULL);

In the future users may wish for larger namespace label sizes.  This
bool option will not allow that.

Perhaps the option should be an integer called "label-size"?

> +static void nvdimm_assert_rw_label_data(NVDIMMDevice *nvdimm, uint64_t size,
> +                                        uint64_t offset)
> +{
> +    assert(nvdimm->reserve_label &&
> +           (nvdimm->label_size >= size + offset) && (offset + size > 
> offset));
> +}

It's not clear from this patch alone, but QEMU is not allowed to assert
due to invalid inputs from the guest.  So if input validation is
necessary here because the values may be invalid, please write if
statements and error returns.

This is important so guests cannot cause QEMU to core dump (SIGABRT
default behavior) and so that nested virtualization doesn't allow a
nested guest to DoS its parent guest.

signature.asc
Description: PGP signature