Re: what's papered over by set_fs(USER_DS) in amd64 signal delivery?

[Ingo Molnar <mingo@elte.hu>]

* Al Viro <viro@ZenIV.linux.org.uk> wrote:

> [...]  IOW, that set_fs() seems to have been useless from the day 1, 
> unless I'm missing something really subtle, like e.g. some processes 
> deliberately running (in 2.0) with %fs set to something with lower 
> limit, with signal handlers allowed to switch back to normal for 
> duration.  And even that would've been broken, since there wouldn't be 
> a matching set_fs() in sigreturn()...

I dont recall us ever having done anything particularly 'clever' with 
in-kernel set_fs()/restore_fs(). Beyond fork/clone it was always 
supposed to be set/restored in a balanced manner. We sometimes leaked it 
unintentionally, and those were security holes.

( Cleverness with security primitives was in fact always actively
  discouraged, even in the early days, as cleverness has the uncanny
  tendency to bit-rot and then has the tendency to slow-convert to a
  security hole by stealth. We always wanted obvious, boringly dumb,
  fail-safe primitives, which can take a few years of bitrot robustly. )

Thanks,

	Ingo