On 11/06/2019 19:18, Volodymyr Babchuk wrote:
Hi Julien,
Hi,
Julien Grall writes:
On 06/06/2019 17:02, Julien Grall wrote:
Hi Volodymyr,
On 5/21/19 10:25 PM, Volodymyr Babchuk wrote:
+static inline bool tee_handle_call(struct cpu_user_regs *regs) +{
- return false;
+}
+static inline int tee_domain_init(struct domain *d, uint16_t tee_type) +{
- return -ENODEV;
+}
I had a report that Xen fails to boot with this series and !CONFIG_TEE. This is because you return an error here in all the case some domain creation will always fail.
Thanks for reporting. I forgot to test that case :(
Instead this should check that tee_type is always NONE or else return an error.
Also, please at least check that your series does not break boot when CONFIG_TEE is not selected. It would also be ideal (but not mandatory) if you can check that it does not break on non-OPTEE platform when !CONFIG_TEE is selected.
I just realized this paragraph may not be clear. What I meant is we need to at least test there are no regression when booting when with CONFIG_TEE=n.
For CONFIG_TEE=y, it would be good to test that it still boots on platform not providing OP-TEE. This is not critical because the config cannot be selected without CONFIG_XEN_EXPERT=y.
I fixed CONFIG_TEE=n issue in the new version, which I'm going to send later today.
Also I made optee_probe() (with CONFIG_OPTEE=y of course) to return false to emulated platform without OP-TEE. System boots and works as usual.
Also I addressed your other comments for this patch.
Thank you! I will have a look at the next version.
Cheers,