On Thu, Aug 06, 2020 at 10:55:43AM -0400, Nathaniel McCallum wrote: > In a past revision of this patch, I had requested a void *misc > parameter that could be passed through vdso_sgx_enter_enclave_t into > sgx_enclave_exit_handler_t. This request encountered some push back > and I dropped the issue. However, I'd like to revisit it or something > similar. > > One way to create a generic interface to SGX is to pass a structure > that captures the relevant CPU state from the handler so that it can > be evaluated in C code before reentry. Precedent for this approach can > be found in struct kvm_run[0]. Currently, however, there is no way to > pass a pointer to such a structure directly into the handler. > > This can be done implicitly by wrapping the struct > sgx_enclave_exception in another structure and then using techniques > like container_of() to find another field. However, this is made more > difficult by the fact that the sgx_enclave_exit_handler_t is not > really using the x86_64 sysv calling convention. Therefore, the > sgx_enclave_exit_handler_t MUST be written in assembly. This also > implies that we can't use techniques like container_of() and must > calculate all the offsets manually, which is tedious, error prone and > fragile. If instead of having a callback, kernel would run an optional bpf program, then a state structure could passed by the kernel to the execution context of the bpf program. That would also sort out difficulty of writing exit handlers. /Jarkko
