Comments for commitment 2 added with '>>>' Hitendra's Commitment review issues 1. Section 6.1.3, need clarification, when domain is restarted after SIR request which CPU will it start? Will this be randomly selected or the same CPU which requested the SIR request will be restarted? RESOLUTION: Spec will be modified to state that the CPU that starts after a SIR is not specified. >>> Restated as determined by platform specific policy. 2. Section 7.1, this section tries to clarify what is cpu list but it does not explain what is cpu id. My expectation from an inception review was that the definition of cpu id would be added to this section. Can we add one of the following texts in section 7 after the second sentence? "The CPU identifiers (cpu id) are assigned by hypervisor via cpu_myid API". Or "The CPU identifiers (cpu id) are part of machine description exported to OBP". RESOLUTION: Spec will be modified to state that the cpu id is a pre-assigned 16 bit value. No further clarification necessary. >>> Confirmed. 3. Section 8.5.7, typo, in the Note section, replace the first "used" with "is". RESOLUTION: Spec will be fixed >>> Confirmed. 4. Section 9.1.1, need clarification, Why scrubbing less memory then client asked for is not an error? RESOLUTION: Closed after discussion Tony Sumpter's commitment review issues 1. 6.1.1 mach_exit It seems that an additional argument to request either restart or "power-off" should be part of this call. Service Entity can over-ride. [Add the extra argument.] RESOLUTION: Closed after discussion about differences between mach_exit and mach_sir 2. 6.1.3 mach_sir "Immediately captured" needs clarification. [Captured as soon as it is possible to interrupt them.] RESOLUTION: Spec will be modified so that the statement says 'as soon as possible' instead of 'immediately'. >>> Confirmed. 3. All explicit and implicit references to OBP and Solaris should be removed, and no dependencies made on documents not referenced in the document. Eg. in the following sections: 7 CPU Services 10 Device interrupt services 10.1 definitions devhandle "hi-cell" and "reg" property are OBP-isms. This doc. is not stated to depend on the Open Firmware bus binding. devino has OBP-isms [Just remove them.] RESOLUTION: Move the 'devino' definition to a 'Note:' to make it informative >>> Confirmed. 4. 8.2 TSB specification Change context to context index (which set of context registers) to match SunSPARC hardware TSB walker design. [Yes.] RESOLUTION: Spec will be modified by project team to implement this suggestion >>> Confirmed. 5. 8.5.1 Maximum of 16 TSBs seems arbitrary: Would prefer a minimum of 2 with a maximum coming from the MD. [Yes.] RESOLUTION: Spec will be modified to state that the maximum number of TSBs will be extracted from the MD (not 16). >>> Confirmed. 6. 8.5.6 Is it invalid to provide a TTE with no valid bit set? [?] RESOLUTION: Spec will be modified to state that if no valid bit is set, behavior is undefined. >>> Confirmed. 7. 8.5.10 mmu_fault_area Is disallowing real address 0 OK? [Yes.] RESOLUTION: Issue is closed after verification. 8. 12.1.2 cons_putchar "currently" needs to be re-worded. [Yes.] RESOLUTION: The word 'Currently' will be removed. >>> Confirmed. 9. 13 Core dump services The mentioned service for the guest to update the buffer is not present. [?] RESOLUTION: Two items need to change: A sentence will be added stating that the buffer will be filled with opaque data by the hypervisor. Also, the name of the command should change to dump_buf_fill. >>> Confirmed, with final name of dump_buf_update. 10. 14.1 ... control structure Who owns the control structure? A malicious or faulty guest must not be able to write it and cause a HV panic. [Make it information only to guest.] RESOLUTION: A sentance will be added to the spec that states something like "The hypervisor owns the control structure and the guest may not modify the data. Attempts to do so will result in undefined behavior." >>> Confirmed. 11. Minor TCRs for to tighten up APIs Provided in file "minor-api-tcrs.txt" [Adopt them.] RESOLUTION: The file has been moved to the commitment materials directory. The project team will update the spec with the changes suggested or will resolve conflicts via e-mail. >>> Confirmed.