blob: c6e8a8dc29d5af2f63fc95ac382167c5583c7e24 (
plain) (
blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
|
Observations:
Questions:
- test/sevstep: why 0 then 15, arent both accesses in the first page?
=> first code page and stack access
- test/sevstep: why does it seem to work with event race but not without?
=> it doesnt (anymore)
Next steps:
- enable single-steping with multiple page faults (only one ends up in
last_fault_gfn and others stay untracked)
- try adjusting timer for single stepping
- Try to setup the non-baseline step without apic, but remapping page
such that we see the relevant page faults and the gfn increment
- test/sevstep: implement counter.. read program memory to see
how many instructions were executed on apic timer
Meeting questions:
- Why is decrypted rip sometimes off-by-one?
=>
- VM gets interrupted independent of our APIC (by the scheduler?) was this a problem for you?
=>
- Is supplementing apic with waiting for next fault viable? (seems to work well actually)
=>
- How does single stepping work with fixed cpu frequency when we always step less than one instruction?
=>
|
