[RFC/Patch] ppc4xx_pm.c

I am in the process of fixing the init of the pm for 4xx in ppc4xx_pm.c
so that it can be used and I and others have discussed having the
drivers enable/disable their own power managment bits at init & exit times.

So what I would like to do is have each core header define a default set
of what is required to be powered on and then have each driver enable
them selves.  These changes would be done in two stages

1) fix ppc4xx_pm.c init and have the default pm bitmap set to all ocp
drivers on.

2) update ocp drivers to enable/disable their pm settings and remove
their enable bits from the default bitmap in the core headers.

armin


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

akuster wrote:

I am in the process of fixing the init of the pm for 4xx in ppc4xx_pm.c
so that it can be used and I and others have discussed having the
drivers enable/disable their own power managment bits at init & exit times.

So what I would like to do is have each core header define a default set
of what is required to be powered on and then have each driver enable
them selves.  These changes would be done in two stages

1) fix ppc4xx_pm.c init and have the default pm bitmap set to all ocp
drivers on.

These patches are the changes to the core header files and for ppc4xx_pm.c.

  Please _note_ that if you enable the Power Managment in the config
that most kernels might hang if you use an ocp device since the drivers
are not updated yet. :)

If no one has an objection, then I will get it submitted

armin

akuster wrote:

I am in the process of fixing the init of the pm for 4xx in ppc4xx_pm.c
so that it can be used and I and others have discussed having the
drivers enable/disable their own power managment bits at init & exit times.

2) update ocp drivers to enable/disable their pm settings and remove
their enable bits from the default bitmap in the core headers.

Here are the ocp driver changes to help support enabling/disabling the
device clock.  currently these wount work on the 440 until a soon to
follow on 440 patch is checked in

armin

 > +++ b/drivers/net/ibm_ocp/ibm_ocp_enet.c
 >

(1) Also do CPM on/off at open/close hooks, so that closed EMACs are
powered down and powered back up when re-opened?

(2) Move ocp_unregister() call outside "if(ocp_get_pm()" test.


--
Todd


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Todd Poynor wrote:

 > +++ b/drivers/net/ibm_ocp/ibm_ocp_enet.c
 >

quoted

     for (curr_emac = 0; curr_emac < emac_max; curr_emac++) {
+        if(ocp_get_pm(EMAC, curr_emac)){
+            mtdcr(DCRN_CPMFR,
+                  mfdcr(DCRN_CPMFR) & ~IBM_CPM_EMAC(ocp_get_pm(EMAC,
curr_emac)));
+        }
         ocp_enet_probe(curr_emac);
     }

@@ -1193,6 +1199,17 @@
 static void __exit
 exit_ppc405_enet(void)
 {
+    int i;
+    struct ocp_dev *emac_dev;
+
+    for (i = 0; i < ocp_get_max(EMAC); i++) {
+        emac_dev = ocp_get_dev(EMAC, i);
+        if(ocp_get_pm(EMAC, i)){
+            mtdcr(DCRN_CPMFR, mfdcr(DCRN_CPMFR) |

IBM_CPM_EMAC(ocp_get_pm(EMAC, i)));
+            ocp_unregister(emac_dev);
+        }
+    }
+

(1) Also do CPM on/off at open/close hooks, so that closed EMACs are
powered down and powered back up when re-opened?

This idea should be applied to all ocp drivers.  I am not sure if
forcing (CPMFR) it the best method @ open/close or requesting/enabling
power down (CPMER).  mmm... gota think about :)

(2) Move ocp_unregister() call outside "if(ocp_get_pm()" test.

Thanks

Armin


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/