Messages in this thread | | | Date | Sat, 31 Jan 2009 01:50:47 +0100 | From | Ingo Molnar <> | Subject | Re: [PATCH] x86: push old stack address on irqstack for unwinder |
| |
* H. Peter Anvin <hpa@zytor.com> wrote:
> Ingo Molnar wrote: >> >> Would be nice to have an #ifdef-less primitive for this - something like: >> >> pushq_frame %rbp >> >> and a matching: >> >> popq_frame %rbp >> >> for those cases that need it (this one doesnt as we dont pop out of the >> stack). >> > > It certainly would if this isn't a singleton, which I think it could > possibly be?
yeah. This is pretty much the only non-restored frame we construct so indeed it would be a singleton. Perhaps the IST ones are such ones too.
> Otherwise it really should be a part of an entry/exit macro; this is > somewhat special in that it sets up a frame pointer as something other > than a normal entry/exit sequence.
Sure - your call really.
Ingo
| |