Message ID | 1433845305-17329-2-git-send-email-noralf@tronnes.org |
---|---|
State | New |
Headers | show |
Hi Noralf, Am 09.06.2015 um 12:21 schrieb Noralf Trønnes: > This adds a new poweroff function to the watchdog driver for the > Raspberry Pi. Currently poweroff/halt results in a reboot. > > [...] > > +static void rpi_power_off(void) > +{ > + struct device_node *np = > + of_find_compatible_node(NULL, NULL, "brcm,raspberrypi-pm-wdt"); > + struct platform_device *pdev = of_find_device_by_node(np); > + struct bcm2835_wdt *wdt = platform_get_drvdata(pdev); > + u32 val; > + > + val = readl_relaxed(wdt->base + PM_RSTS); do you think it's safe here to assume wdt could never be NULL? May be it's necessary to send the series to the watchdog / bcm2835 maintainers to get more feedback. Regards Stefan
On 06/12/2015 04:26 AM, Stefan Wahren wrote: > Hi Noralf, > > Am 09.06.2015 um 12:21 schrieb Noralf Trønnes: >> This adds a new poweroff function to the watchdog driver for the >> Raspberry Pi. Currently poweroff/halt results in a reboot. >> >> [...] >> >> +static void rpi_power_off(void) >> +{ >> + struct device_node *np = >> + of_find_compatible_node(NULL, NULL, "brcm,raspberrypi-pm-wdt"); >> + struct platform_device *pdev = of_find_device_by_node(np); >> + struct bcm2835_wdt *wdt = platform_get_drvdata(pdev); >> + u32 val; >> + >> + val = readl_relaxed(wdt->base + PM_RSTS); > > do you think it's safe here to assume wdt could never be NULL? > If the call is made, the driver must be instantiated. We can therefore assume that neither np, pdev, nor wdt is NULL. If one of those is NULL, it would be a bug, which should not be ignored. Guenter
On 06/09/2015 03:21 AM, Noralf Trønnes wrote: > This adds a new poweroff function to the watchdog driver for the > Raspberry Pi. Currently poweroff/halt results in a reboot. > > The Raspberry Pi firmware uses the RSTS register to know which > partiton to boot from. The partiton value is spread into bits > 0, 2, 4, 6, 8, 10. Partiton 63 is a special partition used by > the firmware to indicate halt. > > The firmware made this change in 19 Aug 2013 and was matched > by the downstream commit: > Changes for new NOOBS multi partition booting from gsh > > Signed-off-by: Noralf Trønnes <noralf@tronnes.org> This poweroff handler stuff is getting really messy :-(. Nothing we can do about that here, so Acked-by: Guenter Roeck <linux@roeck-us.net> Guenter
diff --git a/drivers/watchdog/bcm2835_wdt.c b/drivers/watchdog/bcm2835_wdt.c index 7116968..fdf0d7d 100644 --- a/drivers/watchdog/bcm2835_wdt.c +++ b/drivers/watchdog/bcm2835_wdt.c @@ -36,6 +36,13 @@ #define PM_RSTC_WRCFG_FULL_RESET 0x00000020 #define PM_RSTC_RESET 0x00000102 +/* + * The Raspberry Pi firmware uses the RSTS register to know which partiton + * to boot from. The partiton value is spread into bits 0, 2, 4, 6, 8, 10. + * Partiton 63 is a special partition used by the firmware to indicate halt. + */ +#define PM_RSTS_RASPBERRYPI_HALT 0x555 + #define SECS_TO_WDOG_TICKS(x) ((x) << 16) #define WDOG_TICKS_TO_SECS(x) ((x) >> 16) @@ -159,6 +166,24 @@ static void bcm2835_power_off(void) bcm2835_restart(&wdt->restart_handler, REBOOT_HARD, NULL); } +static void rpi_power_off(void) +{ + struct device_node *np = + of_find_compatible_node(NULL, NULL, "brcm,raspberrypi-pm-wdt"); + struct platform_device *pdev = of_find_device_by_node(np); + struct bcm2835_wdt *wdt = platform_get_drvdata(pdev); + u32 val; + + val = readl_relaxed(wdt->base + PM_RSTS); + val |= PM_PASSWORD | PM_RSTS_RASPBERRYPI_HALT; + writel_relaxed(val, wdt->base + PM_RSTS); + + /* Continue with normal reset mechanism */ + bcm2835_restart(&wdt->restart_handler, REBOOT_HARD, NULL); +} + +static const struct of_device_id bcm2835_wdt_of_match[]; + static int bcm2835_wdt_probe(struct platform_device *pdev) { struct device *dev = &pdev->dev; @@ -192,8 +217,12 @@ static int bcm2835_wdt_probe(struct platform_device *pdev) wdt->restart_handler.notifier_call = bcm2835_restart; wdt->restart_handler.priority = 128; register_restart_handler(&wdt->restart_handler); - if (pm_power_off == NULL) - pm_power_off = bcm2835_power_off; + if (!pm_power_off) { + const struct of_device_id *match; + + match = of_match_node(bcm2835_wdt_of_match, pdev->dev.of_node); + pm_power_off = match->data; + } dev_info(dev, "Broadcom BCM2835 watchdog timer"); return 0; @@ -204,7 +233,7 @@ static int bcm2835_wdt_remove(struct platform_device *pdev) struct bcm2835_wdt *wdt = platform_get_drvdata(pdev); unregister_restart_handler(&wdt->restart_handler); - if (pm_power_off == bcm2835_power_off) + if (pm_power_off == bcm2835_power_off || pm_power_off == rpi_power_off) pm_power_off = NULL; watchdog_unregister_device(&bcm2835_wdt_wdd); iounmap(wdt->base); @@ -218,7 +247,8 @@ static void bcm2835_wdt_shutdown(struct platform_device *pdev) } static const struct of_device_id bcm2835_wdt_of_match[] = { - { .compatible = "brcm,bcm2835-pm-wdt", }, + { .compatible = "brcm,bcm2835-pm-wdt", .data = bcm2835_power_off }, + { .compatible = "brcm,raspberrypi-pm-wdt", .data = rpi_power_off }, {}, }; MODULE_DEVICE_TABLE(of, bcm2835_wdt_of_match);
This adds a new poweroff function to the watchdog driver for the Raspberry Pi. Currently poweroff/halt results in a reboot. The Raspberry Pi firmware uses the RSTS register to know which partiton to boot from. The partiton value is spread into bits 0, 2, 4, 6, 8, 10. Partiton 63 is a special partition used by the firmware to indicate halt. The firmware made this change in 19 Aug 2013 and was matched by the downstream commit: Changes for new NOOBS multi partition booting from gsh Signed-off-by: Noralf Trønnes <noralf@tronnes.org> --- drivers/watchdog/bcm2835_wdt.c | 38 ++++++++++++++++++++++++++++++++++---- 1 file changed, 34 insertions(+), 4 deletions(-)