> On Saturday 17 March 2007, YAMAMOTO Takashi wrote:
> > > http://www.netbsd.org/~xtraeme/msr_ipi_handlers.diff
> >
> > - why iterate over cpus in msr_read_ipi and msr_write_ipi?
> 
> My original patch did not use the second iteration, but making
> many writes to the MSR gave unexpected results and the value
> of the MSRs was not the same.
> 
> Adding the second iteration made it to work all the times I tried.
> 
> Don't tell me why it's needed, but it's needed. If you have the
> information about it, please let me know.

if something behaves in unexpected ways, please investigate why it does.
please don't commit the code which you don't understand.
to me, the iteration seems completely bogus.

> > - why pick the primary cpu in msr_cpu_broadcast_read?
> 
> Isn't it necessary?

you did it because you thought it was necessary for some reasons, didn't you?

> > - what's the point of ci_msr_rvalue?
> 
> Just to know what's the returned value of the MSR in cpuN.

to me, no one in your patch seems to use it in a meaningful way.

YAMAMOTO Takashi