On 03/03/20 18:42, Sean Christopherson wrote: >> As long as we have one use of vcpu, I'd rather skip this patch and >> adjust patch 8 to use "->". Even the other "explicitly take context" >> parts are kinda debatable since you still have to do emul_to_vcpu. >> Throwing a handful of >> >> - struct x86_emulate_ctxt *ctxt = &vcpu->arch.emulate_ctxt; >> + struct x86_emulate_ctxt *ctxt = vcpu->arch.emulate_ctxt; >> >> into patch 8 is not bad at all and limits the churn. > Hmm, I'd prefer to explicitly pass @ctxt, even for the tracepoint. I get > that it's technically unnecessary churn, but explicitly passing @ctxt means > that every funcition that grabs arch.emulate_ctxt (all three of 'em) checks > for a NULL ctxt. That makes it trivial to visually audit that there's no > risk of a bad pointer dereference, and IMO having @ctxt in the prototype > is helpful to see "oh, this helper is called from within the emulator". > That's a good rationale, but I believe this refactoring belongs more in the "disable emulator" part than this one. Paolo
