reset / watchdog on an imx7d soc
Giorgio Dal Molin
giorgio.nicole at arcor.de
Mon Jun 29 12:03:43 EDT 2020
Hi,
> On June 29, 2020 at 5:35 PM Ahmad Fatoum <a.fatoum at pengutronix.de> wrote:
>
>
> Hello Giorgio,
>
> On 6/29/20 5:30 PM, Giorgio Dal Molin wrote:
> > Hi,
> >
> >> On June 29, 2020 at 3:30 PM Ahmad Fatoum <a.fatoum at pengutronix.de> wrote:
> >>
> >>
> >> Hello,
> >>
> >> On 6/29/20 12:53 PM, Giorgio Dal Molin wrote:
> >>> Hi,
> >>>
> >>>> On June 29, 2020 at 10:44 AM Ahmad Fatoum <a.fatoum at pengutronix.de> wrote:
> >>>>
> >>>>
> >>>> Hello Giorgio,
> >>>>
> >>>> On 6/23/20 5:11 PM, Giorgio Dal Molin wrote:
> >>>>> Hi Fabio,
> >>>>>
> >>>>> thank you for your quick reply.
> >>>>>
> >>>>> I've already found the errata you linked in your mail but I had no success
> >>>>> applying the suggestion there; maybe I'm doing it wrong.
> >>>>> Let's take the Option 3 there:
> >>>>
> >>>> Does reset within Linux (with say the imx_v6_v7_defconfg) work?
> >>>
> >>> No. I see the same behavior also with the kernel: after a 'shutdown -r'
> >>> the system 'goes down' but doesn't reboot, it just hang:
> >>>
> >>> ~ # shutdown -r
> >>> system is going down for reboot NOW
> >>> ...
> >>> [ 46.248222] 000: ci_hdrc ci_hdrc.0: remove, state 1
> >>> [ 46.248253] 000: usb usb1: USB disconnect, device number 1
> >>> [ 46.248263] 000: usb 1-1: USB disconnect, device number 2
> >>>
> >>>
> >>> The same happens if I kill the 'watchdog' process: after ~2 seconds
> >>> the system just hangs.
> >>>
> >>> I'm sure there must be a solution because the original u-boot bootloader
> >>> for the cpu module I'm using reboots the system as expected, through the wdog1;
> >>> unfortunately it's not so trivial to find what makes the difference.
> >>
> >> Linux probably attempts reset via PSCI. Check for psci_system_reset
> >> in your vendor's U-Boot and transplant it into barebox.
> >
> > U-Boot has actually a psci_system_reset(), it also rely on wdog1:
> >
> > __secure void psci_system_reset(void)
> > {
> > struct wdog_regs *wdog = (struct wdog_regs *)WDOG1_BASE_ADDR;
> >
> > /* make sure WDOG1 clock is enabled */
> > writel(0x1 << 28, CCM_BASE_ADDR + CCM_ROOT_WDOG);
> > writel(0x3, CCM_BASE_ADDR + CCM_CCGR_WDOG1);
> > writew(WCR_WDE, &wdog->wcr);
> >
> > while (1)
> > wfi();
> > }
> >
> > It is also compiled in the uboot image, because if I add a syntax error in
> > its body the compiler tells me; anyway I think it's not called when I issue
> > the 'reset' command at the prompt because the 'reset' works properly even if
> > I comment out the whole body of the function or if I add an endless loop like
> > while(1); to it.
>
> Yes. It should be called by the kernel when CONFIG_ARM_PSCI_FW=y
> and the device tree has the correct node.
> > Anyway I ported it to barebox, defining a void psci_system_reset(void) in
> > arch/arm/mach-imx/imx7.c:
> >
> > ...
> > #define CCM_BASE_ADDR 0x30380000
> > #define CCM_ROOT_WDOG 0xbb80
> > #define CCM_CCGR_WDOG1 0x49c8
> > #define WCR_WDE 0x04
> > #define WDOG1_WCR 0x30280000
> > #define wfi() __asm__ __volatile__ ("wfi" : : : "memory")
> >
> > static void imx7_system_reset(void)
> > {
> > printf("%s: called now.\n",__func__);
> >
> > // make sure WDOG1 clock is enabled
> > writel(0x1 << 28, CCM_BASE_ADDR + CCM_ROOT_WDOG);
> > writel(0x3, CCM_BASE_ADDR + CCM_CCGR_WDOG1);
> > writew(WCR_WDE, WDOG1_WCR);
> >
> > while (1)
> > wfi();
> > };
> >
> > static struct psci_ops imx7_psci_ops = {
> > .cpu_on = imx7_cpu_on,
> > .cpu_off = imx7_cpu_off,
> > .system_reset = imx7_system_reset,
> > };
> >
> > and the following case to smc.c:
> >
> > static int do_smc(int argc, char *argv[])
> > {
> > ...
> > case 'r':
> > printf("issue psci reset...\n");
> > arm_smccc_smc(ARM_PSCI_0_2_FN_SYSTEM_RESET,
> > 0, 0, 0, 0, 0, 0, 0, &res);
> > printf("reset issued...\n");
> > break;
> > ...
> >
> > to be able to call it.
> > Now, using the modified 'smc' command I can see my printf's:
> >
> > samx7: / smc -n
> > samx7: / smc -r
> > issue psci reset...
> > psci_system_reset called.
> > imx7_system_reset: called now.
> >
> > but the soc still just hangs.
>
> What does U-Boot's reset commando do on an i.MX7?
It ends up calling this function in the imx_watchdog.c:
void __attribute__((weak)) reset_cpu(ulong addr)
{
struct watchdog_regs *wdog = (struct watchdog_regs *)WDOG1_BASE_ADDR;
printf("%s: called now\n",__func__);
udelay(500);
clrsetbits_le16(&wdog->wcr, WCR_WT_MSK, WCR_WDE);
writew(0x5555, &wdog->wsr);
writew(0xaaaa, &wdog->wsr); /* load minimum 1/2 second timeout */
while (1) {
/*
* spin for .5 seconds before reset
*/
}
}
I added the printf / udelay to be sure. WCR_WT_MSK is (0xff << 8).
What I also observed is that if I upload the uboot image with the imx-usb-loader it boots
and runs normally but the reset in this case also just hangs the soc without rebooting it,
like barebox; so there must be a difference between an 'usb uploaded' boot and a 'qspi native'
boot.
giorgio
>
> >
> > giorgio
> >
> >>
> >>>
> >>> giorgio
> >>>
> >>>
> >>>>
> >>>> --
> >>>> Pengutronix e.K. | |
> >>>> Steuerwalder Str. 21 | http://www.pengutronix.de/ |
> >>>> 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
> >>>> Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
> >>>>
> >>>> _______________________________________________
> >>>> barebox mailing list
> >>>> barebox at lists.infradead.org
> >>>> http://lists.infradead.org/mailman/listinfo/barebox
> >>>
> >>
> >> --
> >> Pengutronix e.K. | |
> >> Steuerwalder Str. 21 | http://www.pengutronix.de/ |
> >> 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
> >> Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
> >>
> >> _______________________________________________
> >> barebox mailing list
> >> barebox at lists.infradead.org
> >> http://lists.infradead.org/mailman/listinfo/barebox
> >
>
> --
> Pengutronix e.K. | |
> Steuerwalder Str. 21 | http://www.pengutronix.de/ |
> 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
> Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
>
> _______________________________________________
> barebox mailing list
> barebox at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/barebox
More information about the barebox
mailing list