viafb: don't touch clock state on OLPC XO-1.5
authorDaniel Drake <dsd@laptop.org>
Tue, 4 Sep 2012 15:45:32 +0000 (11:45 -0400)
committerFlorian Tobias Schandinat <FlorianSchandinat@gmx.de>
Sat, 22 Sep 2012 21:31:01 +0000 (21:31 +0000)
As detailed in the thread titled "viafb PLL/clock tweaking causes XO-1.5
instability," enabling or disabling the IGA1/IGA2 clocks causes occasional
stability problems during suspend/resume cycles on this platform.

This is rather odd, as the documentation suggests that clocks have two
states (on/off) and the default (stable) configuration is configured to
enable the clock only when it is needed. However, explicitly enabling *or*
disabling the clock triggers this system instability, suggesting that there
is a 3rd state at play here.

Leaving the clock enable/disable registers alone solves this problem.
This fixes spurious reboots during suspend/resume behaviour introduced by
commit b692a63a.

Signed-off-by: Daniel Drake <dsd@laptop.org>
Signed-off-by: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Cc: stable@vger.kernel.org
drivers/video/via/via_clock.c

index af8f26b643c15cb8510022d53db9144969907697..db1e39277e321f37a39b690d4f6f0d5638325376 100644 (file)
@@ -25,6 +25,7 @@
 
 #include <linux/kernel.h>
 #include <linux/via-core.h>
+#include <asm/olpc.h>
 #include "via_clock.h"
 #include "global.h"
 #include "debug.h"
@@ -289,6 +290,10 @@ static void dummy_set_pll(struct via_pll_config config)
        printk(KERN_INFO "Using undocumented set PLL.\n%s", via_slap);
 }
 
+static void noop_set_clock_state(u8 state)
+{
+}
+
 void via_clock_init(struct via_clock *clock, int gfx_chip)
 {
        switch (gfx_chip) {
@@ -346,4 +351,18 @@ void via_clock_init(struct via_clock *clock, int gfx_chip)
                break;
 
        }
+
+       if (machine_is_olpc()) {
+               /* The OLPC XO-1.5 cannot suspend/resume reliably if the
+                * IGA1/IGA2 clocks are set as on or off (memory rot
+                * occasionally happens during suspend under such
+                * configurations).
+                *
+                * The only known stable scenario is to leave this bits as-is,
+                * which in their default states are documented to enable the
+                * clock only when it is needed.
+                */
+               clock->set_primary_clock_state = noop_set_clock_state;
+               clock->set_secondary_clock_state = noop_set_clock_state;
+       }
 }