[Xen-devel] Re: [patch 13/26] Xen-paravirt_ops: Consistently wrap paravirt ops callsites to make them patchable

[Ingo Molnar <mingo@xxxxxxx>]

* Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:

> I have this dim memory that ARM has done it that way for a long time 
> because it's so expensive to do a "real" cli/sti.
> 
> And I think -rt does it for other reasons. It's just more flexible.

-rt doesnt wrap cli/sti anymore: spin_lock_irq*() doesnt disable irq 
flags on -rt, thus the amount of real irqs-off sections is very small 
and reviewable.

But nevertheless an incarnation of the code survived and is upstream 
already, in the form of TRACE_IRQFLAGS lockdep code ;) This implements a 
soft hardirq flag _today_: all that would be needed is for Xen to define 
raw_local_irq_disable() as a NOP, and to use the 
current->hardirqs_enabled as 'soft IRQ-off flag'.

Note that ->hardirqs_enabled is self-maintained, i.e. it's not just a 
stupid shadow of the hardirq flag, it's an independently maintained flag 
that does not rely on the existence of the hard flag.

[ this code even has its own debugging code, so out-of-sync-flags,
  double-off and double-on is detected and complained about. So all the 
  hard stuff has already been done as part of lockdep, and it's even 
  long-term maintainable because under a native lockdep kernel we check 
  the soft flag against the hard flag. ]

so i think a soft cli/sti flag support should be merged/unified with the 
trace_hardirqs_on()/trace_hardirqs_off() code.

        Ingo

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel