> On Dec 21, 2018, at 11:24 AM, Sean Christopherson > <sean.j.christopher...@intel.com> wrote: > > On Fri, Dec 21, 2018 at 09:12:46AM -0800, Andy Lutomirski wrote: >>> On Dec 21, 2018, at 9:28 AM, Sean Christopherson >>> <sean.j.christopher...@intel.com> wrote: >>> >>> On Wed, Dec 19, 2018 at 06:58:48PM -0800, Andy Lutomirski wrote: >>>>> On Dec 19, 2018, at 6:45 AM, Sean Christopherson >>>>> <sean.j.christopher...@intel.com> wrote: >>>>> >>>>>>> On Wed, Dec 19, 2018 at 09:36:16AM +0000, Jethro Beekman wrote: >>>>> >>>>> I agree with Jethro, passing the enclave_fd as a param is obnoxious. >>>>> And it means the user needs to open /dev/sgx to do anything with an >>>>> enclave fd, e.g. the enclave fd might be passed to a builder thread, >>>>> it shouldn't also need the device fd. >>>>> >>>>> E.g.: >>>>> >>>>> sgx_fd = open("/dev/sgx", O_RDWR); >>>>> BUG_ON(sgx_fd < 0); >>>>> >>>>> enclave_fd = ioctl(sgx_fd, SGX_ENCLAVE_CREATE, &ecreate); >>>>> BUG_ON(enclave_fd < 0); >>>>> >>>>> ret = ioctl(enclave_fd, SGX_ENCLAVE_ADD_PAGE, &eadd); >>>>> BUG_ON(ret); >>>>> >>>>> ... >>>>> >>>>> ret = ioctl(enclave_fd, SGX_ENCLAVE_INIT, &einit); >>>>> BUG_ON(ret); >>>>> >>>>> ... >>>>> >>>>> close(enclave_fd); >>>>> close(sgx_fd); >>>>> >>>>> >>>>> Take a look at virt/kvm/kvm_main.c to see how KVM manages anon inodes >>>>> and ioctls for VMs and vCPUs. >>>> >>>> Can one of you explain why SGX_ENCLAVE_CREATE is better than just >>>> opening a new instance of /dev/sgx for each encalve? >>> >>> Directly associating /dev/sgx with an enclave means /dev/sgx can't be >>> used to provide ioctl()'s for other SGX-related needs, e.g. to mmap() >>> raw EPC and expose it a VM. Proposed layout in the link below. I'll >>> also respond to Jarkko's question about exposing EPC through /dev/sgx >>> instead of having KVM allocate it on behalf of the VM. >> >> Hmm. I guess this makes some sense. My instinct would be to do it a >> little differently and have: >> >> /dev/sgx/enclave: Each instance is an enclave. >> >> /dev/sgx/epc: Used to get raw EPC for KVM. Might have different >> permissions, perhaps 0660 and group kvm. >> >> /dev/sgx/something_else: For when SGX v3 adds something else :) > > Mmmm, I like this approach a lot. It would allow userspace to easily > manage permissions for each "feature", e.g. give all users access to > /dev/sgx/epc but restrict /dev/sgx/enclave. > > And we could add e.g. /dev/sgx/admin if we wanted to exposed ioctls() > that apply to all aspects of SGX. > > Do you know if /dev/sgx/epc could be dynamically created, e.g. by > KVM when the kvm_intel module is loaded?
Presumably sgx would create a bus and enumerate the devices as needed. Or I suppose these things could be platform or system devices. I’m not really a device model expert, and the one time I tried to implement a character device, I got so disgusted that I wrote a whole library for it. It’s still in limbo somewhere. > That would seal the deal for > me as it'd keep open the option of having KVM handle oversubscription > of guest EPC while exposing EPC through /dev/sgx instead of /dev/kvm.