Re: recent change to exit_mmap

[Andrew Morton]

David Mosberger <[e-mail]> wrote:
>
> I don't know why SET_PERSONALITY() came to be where it is now, but it
> does make some sense to me.  One thing that comes to mind: on ia64, we
> normally don't map data segments with execute permission but for
> backwards-compatibility, we need to do that for x86 binaries.  I think
> there might be a problem with that if SET_PERSONALITY() was done too
> late.  Certainly something that could be fixed, but I suspect a
> similar ordering issue (perhaps on SPARC?) might have triggered the
> current placement of SET_PERSONALITY().
> 

hmm.  Seems that all the activities between the two first SET_PERSONALITY()
calls and the flush_old_exec() are pretty innocuous.  And no mappings could
be set up there, because flush_old_exec() would remove them again.

I'll ask Dave about it.