On Mon, Apr 04, 2016 at 08:14:02PM +0200, Geert Uytterhoeven wrote: > Hi Günter, > > On Mon, Apr 4, 2016 at 7:52 PM, Guenter Roeck <linux@xxxxxxxxxxxx> wrote: > > On Mon, Apr 04, 2016 at 07:02:45PM +0200, Geert Uytterhoeven wrote: > >> On Mon, Apr 4, 2016 at 5:25 PM, Guenter Roeck <linux@xxxxxxxxxxxx> wrote: > >> > On Mon, Apr 04, 2016 at 05:21:44PM +0200, Geert Uytterhoeven wrote: > >> >> On Mon, Apr 4, 2016 at 4:59 PM, Wolfram Sang <wsa@xxxxxxxxxxxxx> wrote: > >> >> >> My Salvator-X reboots after timeout from "cat > /dev/watchdog0", but > >> >> >> it doesn't reboot through "reboot" or "reboot -f"? > >> >> > > >> >> > That sadly doesn't work on Gen3. From the RFC v5 cover letter: > >> >> > >> >> > === > >> >> > > >> >> > * drop restart_handler since ARM64 uses PSCI firmware resets which do > >> >> > not call restart handlers > >> >> > > >> >> > The last point was quite a bummer to me because plain reboot was the > >> >> > reason I wrote this driver ;) Well, so is life... > >> >> > >> >> That's indeed silly. Can't we have it as a low-priority restart handler, to > >> > > >> > Yes, it is. It defeats the purpose of restart handlers. PSCI reset should have > >> > been implemented as a high priority restart handler. > >> > >> Unfortunately that won't work: psci_sys_reset() doesn't return on failure. > >> > > You mean it just hangs ? That is bad. If that is the case, it is not reliable > > and thus should be a low priority (or at best medium priority) restart handler > > (which can be replaced with a working higher priority one). > > On my Salvator-X board, it hangs. The call into PSCI doesn't return. > As this is firmware, it may depend on the version of the firmware. > > >> We can still clear arm_pm_restart in platform code, though ;-) > >> > > I had originally planned to replace arm_pm_restart() completely with restart > > handlers. Maybe I should revive the effort ? > > Perhaps. As I remember, it was not such a pretty experience? > That was the attempt to introduce the same mechanism for power-off handling. Yes, that bad experience got me to abandon the attempt to finish the restart handler cleanup as well, but it may still worth at least a final attempt. Guenter
