On 01/11/2015 01:30 PM, Måns Rullgård wrote:
Lars-Peter Clausen <lars@xxxxxxxxxx> writes:
On 01/11/2015 01:15 PM, Måns Rullgård wrote:
Guenter Roeck <linux@xxxxxxxxxxxx> writes:
On 01/10/2015 12:08 PM, Måns Rullgård wrote:
Lars-Peter Clausen <lars@xxxxxxxxxx> writes:
Use the recently introduced do_kernel_restart() function as the default restart
handler if the platform did not explicitly provide a restart handler. This
allows use restart handler that have been registered by device drivers to
restart the machine.
Signed-off-by: Lars-Peter Clausen <lars@xxxxxxxxxx>
---
arch/mips/kernel/reset.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/arch/mips/kernel/reset.c b/arch/mips/kernel/reset.c
index 07fc524..36cd80c 100644
--- a/arch/mips/kernel/reset.c
+++ b/arch/mips/kernel/reset.c
@@ -19,7 +19,7 @@
* So handle all using function pointers to machine specific
* functions.
*/
-void (*_machine_restart)(char *command);
+void (*_machine_restart)(char *command) = do_kernel_restart;
void (*_machine_halt)(void);
void (*pm_power_off)(void);
There is already a similar patch posted by Kevin Cernekee:
http://www.linux-mips.org/archives/linux-mips/2014-12/msg00410.html
Personally I prefer the earlier patch, though I guess that is personal
preference.
They both achieve the same thing, though Kevin's is more in line with
what ARM does. Missing from both is a fallback while(1) loop in case no
restart handlers are registered. With the restart moved to the watchdog
driver, there's a possibility that this might happen.
In my opinion if such a fallback is needed it should be put into the
kernel core reboot implementation and not into individual restart
handler implementations.
My first version of this patch was do_kernel_restart() followed by a
machine_halt() (so it goes to sleep instead of busy looping) as a
fallback. But I couldn't find a good reason why that should be done at
the individual restart handler level, so I dropped it.
ARM does it its arch machine_restart() function. MIPS should probably
follow suit.
That's just cargo culting...
Either it is required, then it should probably be moved into the core
kernel, or it is not required in which case it can be removed from the ARM
machine_restart() implementation.
