On 24/06/20 19:28, Nadav Amit wrote: > Unfortunately this change does not work for me (i.e., it breaks the > task-switch test on my setup). Admittedly, my patch was wrong. > > I actually did not notice this scheme of having the GS descriptor and > GS-base out of sync. It seems very fragile, specifically when you have a > task-switch that reloads the GS and overwrites the original value. > > I will try to investigate it further later. Yeah, it's fragile but it's limited to the task switch tests; and it's the only way to do it without having a separate selector or LDT for each CPU. Paolo
