Changes between Version 2 and Version 3 of Ticket #19171, comment 15
- Timestamp:
- Feb 1, 2020 2:06:31 AM (4 years ago)
Legend:
- Unmodified
- Added
- Removed
- Modified
-
Ticket #19171, comment 15
v2 v3 1 1 We've finally managed to reproduce this issue. The DATEV program as well as CorelDraw X5 are affected. 2 2 3 It seems increasing the VCPU count really helps because it seems to drastically increase the chance of the VMM exiting to ring-3. And the bug (regression) is precisely *only* in the exit-to-ring 3 scenariowhen ICEBP #DB VM-exits occurs.3 It seems increasing the VCPU count really helps because it seems to drastically increase the chance of the VMM exiting to ring-3. And the bug (regression) exits '''only''' in the exit-to-ring 3 case when ICEBP #DB VM-exits occurs. 4 4 5 5 The instruction length of the faulting ICEBP instruction while exiting to ring-3 was incorrectly being reset. Thus while restoring it on re-entry in ring-0, it gets restored as 0xff. Hence VT-x rightly complained that the instruction length is incorrect (should be 0-15).