Re: Reset on Beaglebone Black has become unreliable/broken

[Konstantin Kletschke <konstantin.kletschke@inside-m2m.de>]

On Tue, Jan 07, 2025 at 12:17:03PM +0100, Ahmad Fatoum wrote:

> Thanks for the follow-up.

Of course, my pleasure :)

I used "pio device monitor" from platformio and added a custom filter
searching for ">" in the reveived stream and if true adds
seconds.microseconds timestamp.

With this:

diff --git a/arch/arm/boards/beaglebone/lowlevel.c b/arch/arm/boards/beaglebone/lowlevel.c
index 5dc49dfaaf..db1c464a27 100644
--- a/arch/arm/boards/beaglebone/lowlevel.c
+++ b/arch/arm/boards/beaglebone/lowlevel.c
@@ -97,6 +97,12 @@ extern char __dtb_z_am335x_boneblack_start[];
 extern char __dtb_z_am335x_bone_common_start[];
 extern char __dtb_z_am335x_bone_start[];

+static void __udelay(int us)
+{
+        volatile int i;
+       for (i = 0; i < us * 3; i++);
+}
+
 /**
  * @brief The basic entry point for board initialization.
  *
@@ -111,6 +117,8 @@ static noinline int beaglebone_sram_init(void)
        uint32_t sdram_size;
        void *fdt;

+       int k;
+
        fdt = __dtb_z_am335x_bone_common_start;

        if (is_beaglebone_black())
@@ -134,6 +142,11 @@ static noinline int beaglebone_sram_init(void)
        am33xx_uart_soft_reset((void *)AM33XX_UART0_BASE);
        am33xx_enable_uart0_pin_mux();
        omap_debug_ll_init();
+
+       putc_ll('>');
+       for (k = 0; k < 1000; k++) {
+               __udelay(1000);
+       }
        putc_ll('>');

        barebox_arm_entry(0x80000000, sdram_size, fdt);

This gives me this:

44.412429
>
46.284478
>

If I calculated correct this are 1.8 seconds.
So without the surrounding 1000 loop this final result here:

diff --git a/arch/arm/boards/beaglebone/lowlevel.c b/arch/arm/boards/beaglebone/lowlevel.c
index 5dc49dfaaf..ccae1d1122 100644
--- a/arch/arm/boards/beaglebone/lowlevel.c
+++ b/arch/arm/boards/beaglebone/lowlevel.c
@@ -97,6 +97,12 @@ extern char __dtb_z_am335x_boneblack_start[];
 extern char __dtb_z_am335x_bone_common_start[];
 extern char __dtb_z_am335x_bone_start[];

+static void __udelay(int us)
+{
+        volatile int i;
+       for (i = 0; i < us * 3; i++);
+}
+
 /**
  * @brief The basic entry point for board initialization.
  *
@@ -135,6 +141,7 @@ static noinline int beaglebone_sram_init(void)
        am33xx_enable_uart0_pin_mux();
        omap_debug_ll_init();
        putc_ll('>');
+       __udelay(1000); // Adding delay of 1.8ms

        barebox_arm_entry(0x80000000, sdram_size, fdt);
 }

Should add a delay of 1.8ms. Is this feasible?


> If it's indeed just a millisecond, I am in favor of adding this to
> barebox upstream with a comment explaining what this fixes, even
> if we don't fully understand it yet.

Yes, I agree to the not fully yet understood part.
What me holds back from lowering the speed is the following:

I went into production with the final result part above and reflashe all
our BBB devices which failed warm start without the additional delay and
all passed 100% flawlessly after the modficication. So it is tested upon
roughly 30 devices which failed before.

Kind Regards
Konstantin



-- 
INSIDE M2M GmbH
Konstantin Kletschke
Berenbosteler Straße 76 B
30823 Garbsen

Telefon: +49 (0) 5137 90950136
Mobil: +49 (0) 151 15256238
Fax: +49 (0) 5137 9095010

konstantin.kletschke@inside-m2m.de
http://www.inside-m2m.de 

Geschäftsführung: Michael Emmert, Derek Uhlig
HRB: 111204, AG Hannover