cvs commit: src/lib/libthr/arch/alpha/alpha _curthread.c

[Andrew Gallatin]

Marcel Moolenaar [marcel at xcllnt.net] wrote:
> On Thu, Jul 24, 2003 at 05:00:13PM -0400, John Baldwin wrote:
> > 
> > There is a chicken and egg problem.  We use the unique value to
> > initialize the per-cpu pointer on kernel entry.  We only have one
> > such beast, and in the kernel we cache it in a register that userland
> > gets to clobber (and frequently does).  If you can think of a better
> > way to store the per-cpu pointer such that kernel entry can load it
> > into the register go for it.
> 
> I see where the confusion is:
> 
> >         pcpup = (struct pcpu *) alpha_pal_rdval();
> 				^^^^^^^^^^^^^^^^^^^^^^
> 
> This is not the per-CPU unique value. I had to add alpha_pal_rdunique()
> in order to access it. In numbers:
> 
> PAL_rdval = PAL_OSF1_rdval = 0x0032
> PAL_rdunique = 0x009e
> 
> PAL_rdval is a privileged operation PAL_rdunique isn't.
> 
> There is no fundamental problem (although my alpha does reboot
> when I run a 1:1-threaded application, so there is a problem :-)
> 

FWIW, if we need another way to get to pcpu, we can always use
alpha_pal_whami() to index into an array of pre-allocated per-cpu
structs. 

Drew