Re: svn commit: r225892 - head/sys/mips/mips

2011-10-01 Thread Alexander Motin
Adrian Chadd wrote: > On 1 October 2011 15:21, Jayachandran C. wrote: > >> I would like to understand this, reverting r216862 would take out the >> critical_enter() and cpu_idleclock() which would keep the timer >> interrupts coming is as usual, this should not affect the event timer. > > I have

Re: svn commit: r225892 - head/sys/mips/mips

2011-10-01 Thread Alexander Motin
Adrian Chadd wrote: > .. and somehow linux mips code does do a variety of WAIT-y things; how > is it they don't have the interrupt handling issues we do? > Is it because they're doing preemption? If so, how do they accurately > handle hz clock pulses when an interrupt may preempt things just > befo

Re: svn commit: r225892 - head/sys/mips/mips

2011-10-01 Thread Adrian Chadd
.. and somehow linux mips code does do a variety of WAIT-y things; how is it they don't have the interrupt handling issues we do? Is it because they're doing preemption? If so, how do they accurately handle hz clock pulses when an interrupt may preempt things just before that wait instruction occur

Re: svn commit: r225892 - head/sys/mips/mips

2011-10-01 Thread Adrian Chadd
On 1 October 2011 15:21, Jayachandran C. wrote: > I would like to understand this, reverting r216862 would take out the > critical_enter() and cpu_idleclock() which would keep the timer > interrupts coming is as usual, this should not affect the event timer. Because the fundamental problem still

Re: svn commit: r225892 - head/sys/mips/mips

2011-10-01 Thread Jayachandran C.
On Sat, Oct 1, 2011 at 12:39 PM, Adrian Chadd wrote: > Reverting it unfortunately doesn't help; it messes up the event timer stuff. I would like to understand this, reverting r216862 would take out the critical_enter() and cpu_idleclock() which would keep the timer interrupts coming is as usual,

Re: svn commit: r225892 - head/sys/mips/mips

2011-10-01 Thread Adrian Chadd
Reverting it unfortunately doesn't help; it messes up the event timer stuff. I'd like to see a more permanent solution. What I suggest we do moving forward is: * create a function pointer setup, like what i386 does * override it per-platform and per-chipset, once we know what the correct behaviou

Re: svn commit: r225892 - head/sys/mips/mips

2011-09-30 Thread Jayachandran C.
Hi Adrian, On Sat, Oct 1, 2011 at 11:26 AM, Adrian Chadd wrote: > Author: adrian > Date: Sat Oct  1 05:56:25 2011 > New Revision: 225892 > URL: http://svn.freebsd.org/changeset/base/225892 > > Log: >  Disable using wait in cpu_idle() until a better solution to timer and >  interrupt handling can

svn commit: r225892 - head/sys/mips/mips

2011-09-30 Thread Adrian Chadd
Author: adrian Date: Sat Oct 1 05:56:25 2011 New Revision: 225892 URL: http://svn.freebsd.org/changeset/base/225892 Log: Disable using wait in cpu_idle() until a better solution to timer and interrupt handling can be implemented. Modified: head/sys/mips/mips/machdep.c Modified: head/sys/m