[PATCH] rtc: zynqmp: Enable crystal oscillator in initialization

[Alexandre Belloni]

On 24/09/2026 09:19:16+0200, Daniel Viaño wrote:
> Enable the RTC crystal oscillator bit in the control register (RTC_CTRL,
> offset 0x40) during driver initialization.
> 
> Although the RTC_OSC_EN macro was defined when the driver was originally
> introduced, it was never asserted in xlnx_init_rtc(). On cold boots, after
> battery exhaustion, or on platforms where firmware does not pre-initialize
> RTC_CTRL, the oscillator remains disabled and the counter never ticks.
> 
> Asserting RTC_OSC_EN alongside RTC_BATT_EN ensures the oscillator starts
> without relying on bootloader initialization. Per the Zynq UltraScale+
> TRM (UG1085), bit 24 (OSC_CNTRL) is a static level enable for the crystal
> inverter rather than an edge-triggered reset; re-asserting it when the
> oscillator is already running is a non-destructive no-op that does not
> glitch or reset the ticking counters.
> 
> Tested on Zynq UltraScale+ hardware, confirming that the counter freezes
> when RTC_OSC_EN is cleared, advances reliably once enabled, and is
> unaffected when re-asserted while running.
> 
> Fixes: 11143c19eb57 ("rtc: add xilinx zynqmp rtc driver")
> Cc: stable at vger.kernel.org
> Signed-off-by: Daniel Viaño <danividanivi at gmail.com>
> ---
>  drivers/rtc/rtc-zynqmp.c | 7 +++++--
>  1 file changed, 5 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/rtc/rtc-zynqmp.c b/drivers/rtc/rtc-zynqmp.c
> index 5bcb7536e973..278705ec09f5 100644
> --- a/drivers/rtc/rtc-zynqmp.c
> +++ b/drivers/rtc/rtc-zynqmp.c
> @@ -167,9 +167,12 @@ static void xlnx_init_rtc(struct xlnx_rtc_dev *xrtcdev)
>  {
>  	u32 rtc_ctrl;
>  
> -	/* Enable RTC switch to battery when VCC_PSAUX is not available */
> +	/*
> +	 * Enable crystal oscillator, and enable RTC switch to battery
> +	 * when VCC_PSAUX is not available.
> +	 */
>  	rtc_ctrl = readl(xrtcdev->reg_base + RTC_CTRL);
> -	rtc_ctrl |= RTC_BATT_EN;
> +	rtc_ctrl |= RTC_BATT_EN | RTC_OSC_EN;

No, you need to keep it disabled until you now the time is correct, that
is in set_time and you need to test for it in read_time and return
-EINVAL when it is not set.


-- 
Alexandre Belloni, co-owner and COO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com