[FAQ] AM5748: AM57XX: MPU COUNTER_REALTIME saturates after several hundred days

The errata document: https://www.ti.com/lit/er/sprz429m/sprz429m.pdf?ts=1633078446981&ref_url=https%253A%252F%252Fwww.google.com%252F

Page: 94

Describes the possible way to avoid & Workaround the issue under discussion.

Workaround

1)  Avoid the issue by requiring the user to preform a scheduled reset/restart of the system sometime prior the saturation time. Workaround

2)  Recover from the issue by enabling a system watchdog reset to expire after the issue has occurred. Workaround

3) Use GP Timer instead of COUNTER_REALTIME in systems requiring continuous up time.

For Systems that cannot be reset before the COUNTER_REALTIME expiry the only option is to switch to GP Timer instead.

The upstream backport of patches on 4.19 Kernel which can be applied on 6.3 Linux AM57xx SDK:
https://patchwork.kernel.org/project/linux-omap/list/?series=512953

The zip of all the 4 patches in the above link that are needed to switch the clock source to GP Timers here:

4.19-counter_realtime_backport.zip

5.4 Kernel backports:

https://patchwork.kernel.org/project/linux-omap/patch/20210602104625.6079-1-tony@atomide.com/
https://patchwork.kernel.org/project/linux-omap/patch/20210602104625.6079-2-tony@atomide.com/
https://patchwork.kernel.org/project/linux-omap/patch/20210602104625.6079-3-tony@atomide.com/
https://patchwork.kernel.org/project/linux-omap/patch/20210602104625.6079-4-tony@atomide.com/

5.10 kernel backports: 

http://lkml.iu.edu/hypermail/linux/kernel/2105.2/00585.html
http://lkml.iu.edu/hypermail/linux/kernel/2105.2/00586.html

Note: Now that we switch from COUNTER_REALTIME to GP Timer to work around the errata the COUNTER_REALTIME is NO longer initialized.
So a quite common solution to read the timers to get a quick and efficient timer read:

Uint32 nowl = 0, nowu = 0;
asm volatile("mrrc p15, 1, %0, %1, c14" : "=r"(nowl), "=r"(nowu));