Re: port/sparc64 known working on Netra X1?

[Chris Ross <cross+netbsd%distal.com@localhost>]

On May 8, 2010, at 20:00, Eduardo Horvath wrote:
> Psycho chips have registers that can generate interrupts for things  
> such
> as power failure if they are wired up to the proper sensors.  It may  
> be
> that the X1 has those things floating and generates spurious  
> interrupts.
> Can you do a .properties on the top level PCI device node?  And  
> maybe put
> a printf in psycho_powerfail() to check if it triggers.

  You are quite right.  The issue is not that it hangs, but that it  
generates a huge, if not unending, flood of interrupts.  After putting  
a printf() at the top of psycho_powerfail(), I see:


** done; installing intr for psycho_cw
@@ malloc'ing 72 bytes
@@ got 0x2789400
@@ calling intr_establish()
@@ back from intr_establish(), ih_map is 0xe0023078, and contains  
0x00000000000007ef.  Going to OR in 0x0000000080000000
@@ returning from psycho_set_intr
** done; installing intr for psycho_bus_a
@@ malloc'ing 72 bytes
@@ got 0x2789380
@@ calling intr_establish()
@@ back from intr_establish(), ih_map is 0xe0023080, and contains  
0x00000000000007f0.  Going to OR in 0x0000000080000000
@@ returning from psycho_set_intr
** done; installing intr for psycho_powerfail
@@ malloc'ing 72 bytes
@@ got 0x2789300
@@ calling intr_establish()
@@ back from intr_establish(), ih_map is 0xe0023028, and contains  
0x00000000000007e5.  Going to OR in 0x0000000080000000
%% in psycho_powerfail
%% in psycho_powerfail
%% in psycho_powerfail
%% in psycho_powerfail
%% in psycho_powerfail
[...]

  Which just goes on forever.

  Does this help determine what may be the "right" solution?

            - Chris