On Thu, 30 Mar 2000, Richard Henderson wrote:

>Ug.  This would prevent the bootloader from writing interesting data
>into bss addresses (after having cleared the thing of course).

The bss section have to be _zeroed_ (IA32 is doing this correctly for
ages, and yes after the bootloader exited). If something that has not to
be zero is allocated in the bss section that's a plain _bug_ in the
bootloader<->kernel interface.

The bootloader must not load _anything_ in the bss section. Magic
variables that deals with the bootloader _have_ to remains separate (at
hardwired offsets) for robusteness. And infact they are always been this
way and the bss-clear-in-common-code patch works fine.

Andrea


Reply via email to