I guess that makes sense. Should I submit a patch for the same as you
suggested?

On Wed, Jul 25, 2018, 11:15 PM Maxime Villard <[email protected]> wrote:

> Le 25/07/2018 à 19:33, Siddharth Muralee a écrit :
> > KERN_BASE is used only once in dtrace and no where else - if it indeed is
> > the same as VM_MIN_KERNEL_ADDRESS then we can just substitute it.
> > I didn't understand how VM_MAX_ADDRESS comes there.
>
> Instead of taking the kernel min address, you can take the userland max
> address. This userland max address is VM_MAX_ADDRESS.
>
> > Also why is KERNBASE > VM_MAX_KERNEL_ADDRESS in amd64 and equal to
> > VM_MIN_KERNEL_ADDRESS in i386(or did you mean VM_MAX_KERNEL_ADDRESS)
>
> I did mean VM_MIN_KERNEL_ADDRESS. The reason is that on amd64 we are
> required to put the kernel at the last 2GB of the VA space, because of the
> mcmodel. So we have KERNBASE that is near the end of the VA space, and
> VM_MIN_KERNEL_ADDRESS that is located before. The two areas are separate.
>
> On i386 however we can put the kernel wherever we want, and by default we
> chose to put it at VM_MIN_KERNEL_ADDRESS. So there, VM_MIN_KERNEL_ADDRESS
> equals KERNBASE.
>

Reply via email to