TSM capabilities and minor editorial changes #82
Conversation
Signed-off-by: Wojciech Ozga <[email protected]>
Signed-off-by: Wojciech Ozga <[email protected]>
…memory and lack of dynamic page conversion. Signed-off-by: Wojciech Ozga <[email protected]>
…ng in a more precise way what MMIO regions must be removed. Signed-off-by: Wojciech Ozga <[email protected]>
…s as a negative value (-1) but declared arguments as unsigned Signed-off-by: Wojciech Ozga <[email protected]>
Signed-off-by: Wojciech Ozga <[email protected]>
Signed-off-by: Wojciech Ozga <[email protected]>
Signed-off-by: Wojciech Ozga <[email protected]>
Signed-off-by: Wojciech Ozga <[email protected]>
Signed-off-by: Wojciech Ozga <[email protected]>
…entation IDs Signed-off-by: Wojciech Ozga <[email protected]>
…entation IDs Signed-off-by: Wojciech Ozga <[email protected]>
src/sbi_cove.adoc
Outdated
| /* | ||
| * CoVE features supported by the running TSM. | ||
| * Enabled bit `i` indicates that the TSM and underly supports capability `i` and there | ||
| * is hardware support. |
There was a problem hiding this comment.
The comment sounds bit ambiguous to me w.r.t "there is hardware support". I think something along the lines is sufficient.
"A bitmask of CoVE features supported by the running TSM." - We can refer to the below capability table for additional clarification.
There was a problem hiding this comment.
The reason why I wrote it like this is that there are cases when software (TSM) and hardware support different set of features. The TSM must report to OS/VMM only features that it can use, i.e., the ones that are supported by software and hardware.
For example, take a platform with hardware support for AIA. If someone runs there a TSM that does not implement support for AIA, OS/VMM would have to use only TSM's legacy interrupt handling.
There was a problem hiding this comment.
I added a reference to the table as suggested.
|
Apart from those nit comments, looks good to me. |
Signed-off-by: Wojciech Ozga <[email protected]>
…nd RV128 Signed-off-by: Wojciech Ozga <[email protected]>
Signed-off-by: Wojciech Ozga <[email protected]>
Signed-off-by: Wojciech Ozga <[email protected]>
|
Thanks Atish and Ravi for feedback and valuable comments. I have addressed them and pushed already changes. |
We propose to extend the COVH get_tsm_info() call with additional information called
TSM capabilitiesto simplify implementation of a hypervisor that supports different CoVE deployment models. Specifically, TSM informs the hypervisor via thetsm_infostructure what capabilities is supports, such as support for single- or mult-step TVM creation, local and/or remote attestation, legacy interrupt handling or AIA, static or dynamic memory allocation.Additionally, we provide a set of minor editorial changes that clarify: (1) how interrupts are injected if AIA is not supported, (2) what properties local attestation can offer for embedded platforms, (3) how TVM enables or disables injection of all interrupts (wildcard), (4) how MMIO regions are specified, (5) how TVM retrieves local attestation secret.