In article <46908B5C.7020306@netbsd.org>
kardel@NetBSD.org wrote:

> I can check x86 again on AMD X2 most likely next WE.

Thanks, I've also put compiled GENERIC.MP kernels:
http://www.ceres.dti.ne.jp/~tsutsui/netbsd/netbsd-alpha-GENERIC.MP-cctr-20070707.gz
http://www.ceres.dti.ne.jp/~tsutsui/netbsd/netbsd-amd64-GENERIC.MP-cctr-20070707.gz
http://www.ceres.dti.ne.jp/~tsutsui/netbsd/netbsd-i386-GENERIC.MP-cctr-20070707.gz
http://www.ceres.dti.ne.jp/~tsutsui/netbsd/alpha-cctr-todr-20070707.diff

> Maybe we find the problem in the cctr code that leads to
> backward time steps on MP systems detected by
> regress/sys/kern/time.

I'm not familiar with MP issue, but splhigh()/splx() pairs should
be replaced with some proper locks?
---
Izumi Tsutsui

> Izumi Tsutsui wrote:
> > In article <45E03427.5080304@netbsd.org>
> > kardel@NetBSD.org wrote:
> >
> >   
> >> Izumi Tsutsui wrote:
> >>     
> >
> >   
> >>> I'm not sure what is the best way to switch alpha to timecounter,
> >>> but I just take functions from x86/x86/tsc.c and adapt it to
> >>> alpha's PCC counter.
> >>>
> >>> Could anyone review the attached patch?
> >>> (which also includes generic todr(9) changes)
> >>>       
> >  :
> >   
> >> In general I agree with the strategy but I think we
> >> should try to avoid a copy if tsc.c here. I include an
> >> old patch here that attempts to split tsc.c into a
> >> MD part (tsc.c) and an MI part kern_cctr.c. Maybe you can
> >> use that for the cycle counter part of the part - note on
> >> alpha this patch lacks the cc_init() call. Also note
> >> that the patch assumes that the counter is read by
> >> cpu_counter32() - same headfile magic as was used
> >> for kern_microtime.c.
> >>     
> >
> > I've updated my todr(9) and timecounter(9) patch for alpha
> > based on your MI kern_cctr.c patch (attached):
> >
> > - use todr(9) API with MI mc146818(4) driver and remove homegrown
> >   todr stuff from MD alpha/clock.c and alpha/mcclock.c
> > - also remove cc_microtime stuff from MD code
> > - add ci_pcc_freq member in struct cpu_info for cpu_frequency(), and
> >   calibrate it with mc146818 interval clock in mcclock attachment
> > - tsc.c is split into MI kern_cctr.c and MD tsc.c as kardel's patch
> >   (with some cosmetics)
> > - call cc_init() in cpu_initclocks(9) because all alpha cpus have
> >   a pcc counter
> > - prepare new <sys/ctrr.h> which is just copied from
> >   __HAVE_TIMECOUNTER part of <sys/cc_microtime.h> so that
> >   we can migrate alpha and x86 to kern_cctr.c independently
> >   (the attached patch also contains changes for i386 and amd64)
> >
> > Tested on AlphaPC164 366MHz and at least todr(9) and ntpd(8)
> > work fine. I'll also test it on DEC3000/300 and Duron PC,
> > but I don't have any SMP machines.