On 10/25/2011 11:27 AM, Paul Brook wrote: >> - /* Avoid mode switch when restoring CPSR. */ >> - env->uncached_cpsr = val & CPSR_M; >> - cpsr_write(env, val, 0xffffffff); >> + >> + env->uncached_cpsr = env->cpsr_vmstate & CPSR_M; >> + cpsr_write(env, env->cpsr_vmstate, 0xffffffff); > > You've removed the helpful comment explaining why this wierdness exists. > Please put it back. > > A comment in cpu.h explaining what cpsr_vmstate is for would also be nice. > Specifically it's a hack[1] to transfer data between the pre_save/post_load > hooks and the vmstate machinery. IMO "vmstate" is a sufficiently generic > term > that it could be confused with some actual cpu register, especially once we > get round to implementing the hypervisor extensions. > > Paul > > [1] Maybe a necessary hack, but still a wart from requiring static table > driven vmstate descriptions.
I wonder if we shouldn't make it easier to do what I do for Alpha with the rather different hack for the fpcr. That might avoid some of the other pre/post_load hooks, and the extra data fields that requires in the cpu structure. r~
