Re: [RFC] Make __bss_start point at actual start of .bss

[Michael Matz <matz@suse.de>]

Hello,

On Mon, 26 Jun 2023, Andreas Krebbel via Binutils wrote:

> > The question is, how is __bss_start used?  Is it the start of the .bss
> > section or the start of the bss segment?  I think it is mainly used as
> > the start of the segment, when loaders zero out all bss memory, so it
> > should stay before other bss-style sections like .sbss.
> 
> The usage which triggered the problem came from qemu and the code there 
> clears memory from __bss_start through _end. This should work.

But that's _exactly_ the case that Alan worried about: __bss_start being 
assumed to be the start of the bss segment, not merely of the section 
named .bss.  If bss-style sections come before .bss (like .sbss does on 
some targets) then __bss_start should be in front of those ones, otherwise 
the above clearing that qemu does would not work anymore (.sbss wouldn't 
be cleared).  See here (elf.sc):

  ${RELOCATING+${CREATE_SHLIB+PROVIDE (}${USER_LABEL_PREFIX}__bss_start = .${CREATE_SHLIB+)};}
  ${RELOCATING+${OTHER_BSS_SYMBOLS}}
  ${DATA_SDATA-${SBSS}}
  ${BSS_PLT+${PLT}}
  .${BSS_NAME}          ${RELOCATING-0} :
  {
     ... normal bss stuff ...

__bss_start needs to remain in front of the SBSS entry for the 
traditional usage to keep working (on targets where DATA_SDATA isn't set).


Ciao,
Michael.