Subject: Re: misc/37612: cpu_switchto(9) doesn't document 'returning' parameter
To: None <,,>
From: Andrew Doran <>
List: netbsd-bugs
Date: 12/25/2007 22:00:04
The following reply was made to PR misc/37612; it has been noted by GNATS.

From: Andrew Doran <>
Subject: Re: misc/37612: cpu_switchto(9) doesn't document 'returning' parameter
Date: Tue, 25 Dec 2007 21:57:54 +0000

 On Wed, Dec 26, 2007 at 08:36:59AM +1100, matthew green wrote:
 >    Architecture: x86_64
 >    Machine: amd64
 >    >Description:
 >    	The cpu_switchto(9) man page declares the function as:
 It's on my list of pages to update; I haven't gotten around to it yet.
 >    	    lwp_t *cpu_switchto(lwp_t *oldlwp, lwp_t *newlwp);
 >    	But should now read:
 >    	    lwp_t *cpu_switchto(lwp_t *oldlwp, lwp_t *newlwp, bool returning);
 > interesting...
 > andy, what is this bool returning?  i notice that most
 > platforms cpu_switchto() don't notice it..
 It can safely be ignored if __HAVE_FAST_SOFTINTS is not defined.
 From -
   Minor changes to cpu_switchto(). It takes a extra boolean argument,
   'returning'.  If returning is set, it means we are returning for a soft
   interrupt that is blocking (so will resume in the trampoline function). 
   If it's set, it's a good indication that any kind of address space or
   user activity can be completely ignored (e.g. activity around
   ras_lookup(), cache flushes, TLB wirings, lazy FPU state, etc). All
   that's needed is to restore the register state and stack then return;
   everything else has been 'borrowed' from the interrupted LWP that we are
   returning to.