From 82d56ef77c114ac0b375fef04cea3a50f10f1843 Mon Sep 17 00:00:00 2001 From: Louis Burda Date: Tue, 22 Nov 2022 15:03:02 +0100 Subject: Improved single stepping and added initial cpuid events --- notes | 28 ---------------------------- 1 file changed, 28 deletions(-) delete mode 100644 notes (limited to 'notes') diff --git a/notes b/notes deleted file mode 100644 index c6e8a8d..0000000 --- a/notes +++ /dev/null @@ -1,28 +0,0 @@ -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? - => - - -- cgit v1.2.3-71-gd317