Messages in this thread | | | Date | Tue, 5 Feb 2008 18:33:58 +0100 | From | Pavel Machek <> | Subject | Re: brk randomization breaks columns |
| |
On Tue 2008-02-05 08:58:41, Arjan van de Ven wrote: > On Tue, 5 Feb 2008 16:46:48 +0100 > Pavel Machek <pavel@ucw.cz> wrote: > > > Hi! > > > > > > In my usual dither, I'm rather hoping Arjan will have a clear > > > > answer. > > > > > > > > > setarch works. If the apps come in source form they need fixing > > > anyway (since I'd not be surprised of current gcc reorders > > > variables), if not.. we only have 2 cases, the other case was the > > > build process of emacs (which got fixed 5 years ago). > > > > uemacs ... broken with randomization > > colums, sss ... local programs, broken with randomization > > procinfo ... broken, randomization makes it die sooner. > > mikmod ... broken with randomization > > bsdsed ... broken with randomization > > ... > > Should I test few more? > > ok throw that idea out of the window then > > the combo of a config option + sysctl sounds the right way forward then ;( > > Unless there's a way we can make sys_brk() detect this kind of behavior somehow... > we could track per process if brk(0) was called, and if not, do something fancy to work around stuff?
That behaviour is indeed easy to detect. If someone does sys_brk() with value lower than heap_start, it means it is old binary (but I'd call that a hack).
But... older binaries call sys_personality(). Maybe it is cleaner not to base check on that? Pavel -- (english) http://www.livejournal.com/~pavelmachek (cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
| |