Re: [PATCH 00/14] selftests/sgx: Oversubscription, page permission, thread entry

[Jarkko Sakkinen <jarkko@kernel.org>]

On Wed, 2021-09-15 at 13:30 -0700, Reinette Chatre wrote:
> Hi Everybody,
> 
> This series consists out of outstanding SGX selftests changes, rebased
> and gathered in a single series that is more easily merged for testing
> and development, and a few more changes added to expand the existing tests.
> 
> The outstanding SGX selftest changes included in this series that have already
> been submitted separately are:
> 
> * An almost two year old patch fixing a benign linker warning that is still
>   present today:
>   https://lore.kernel.org/linux-sgx/20191017030340.18301-2-sean.j.christopherson@intel.com/
>   The original patch is added intact and not all email addresses
>   within are valid.
> 
> * Latest (v4) of Jarkko Sakkinen's series to add an oversubscription test:
>   https://lore.kernel.org/linux-sgx/20210809093127.76264-1-jarkko@kernel.org/
> 
> * Latest (v2) of Jarkko Sakkinen's patch that provides provide per-op
>   parameter structs for the test enclave:
>   https://lore.kernel.org/linux-sgx/20210812224645.90280-1-jarkko@kernel.org/
> 
> The reason why most of these patches are outstanding is that they depend
> on a kernel change that is still under discussion. Decision to wait in:
> https://lore.kernel.org/linux-sgx/f8674dac5579a8a424de1565f7ffa2b5bf2f8e36.camel@kernel.org/
> The original patch for this kernel dependency continues to be included in
> this series as a placeholder until the ongoing discussions are concluded.
> 
> The new changes introduced in this series builds on Jarkko's outstanding
> SGX selftest changes and adds new tests for page permissions, exception
> handling, and thread entry.
> 
> Reinette
> 
> Jarkko Sakkinen (9):
>   x86/sgx: Add /sys/kernel/debug/x86/sgx_total_mem
>   selftests/sgx: Assign source for each segment
>   selftests/sgx: Make data measurement for an enclave segment optional
>   selftests/sgx: Create a heap for the test enclave
>   selftests/sgx: Dump segments and /proc/self/maps only on failure
>   selftests/sgx: Encpsulate the test enclave creation
>   selftests/sgx: Move setup_test_encl() to each TEST_F()
>   selftests/sgx: Add a new kselftest: unclobbered_vdso_oversubscribed
>   selftests/sgx: Provide per-op parameter structs for the test enclave
> 
> Reinette Chatre (4):
>   selftests/sgx: Rename test properties in preparation for more enclave
>     tests
>   selftests/sgx: Add page permission and exception test
>   selftests/sgx: Enable multiple thread support
>   selftests/sgx: Add test for multiple TCS entry
> Sean Christopherson (1):
>   selftests/x86/sgx: Fix a benign linker warning
> 
>  Documentation/x86/sgx.rst                     |   6 +
>  arch/x86/kernel/cpu/sgx/main.c                |  10 +-
>  tools/testing/selftests/sgx/Makefile          |   2 +-
>  tools/testing/selftests/sgx/defines.h         |  33 +-
>  tools/testing/selftests/sgx/load.c            |  40 +-
>  tools/testing/selftests/sgx/main.c            | 341 +++++++++++++++---
>  tools/testing/selftests/sgx/main.h            |   7 +-
>  tools/testing/selftests/sgx/sigstruct.c       |  12 +-
>  tools/testing/selftests/sgx/test_encl.c       |  60 ++-
>  .../selftests/sgx/test_encl_bootstrap.S       |  21 +-
>  10 files changed, 445 insertions(+), 87 deletions(-)
> 

One test that would be also nice to have at some point would
be vepc test. It's not exceptionally hard to ramp up KVM:

https://lwn.net/Articles/658511/

Hmm... perhaps this type of kselftest should be part of the
series that Paolo is upstreaming because otherwise we are
dependent on non-upstream QEMU to test those changes.

Looking back, this would have been already good idea to ramp
up when the original KVM-SGX series was upstreamed because not
that many have motivation to self-compile QEMU (I did but thinking
about potential larger coverage).

/Jarkko