U-Boot code from current BSP (lf-6.6.52_2.2.0) contains a fix for an undocumented erratum. Please give more in detail information about the issue. Is there a fix available for Linux?
Hi @Niebel-TQ
Is this always guaranteed even for a kernel compiled with PREEMPT_RT and / or at higher tick rates?
-->The Linux OS here is NONE-RT kernel, for NONE-RT kernel, we have tested and didn't observe issue. For PREEMPT_RT Linux, there is no test data on this, but according to previous information, for Real Time OS, this delay needs to be added.
Is this issues limited to i.MX93 or is it present in other CPU / MCU
-->The affected chips are i.MX91, i.MX93 and i.MX95 A1.
Best Regards,
Zhiming
Thank you for explaining.
Two additiponal questions:
1) `In Linux OS, since software latency is enough`: Is this always guaranteed even for a kernel compiled with PREEMPT_RT and / or at higher tick rates?
2) Is this issues limited to i.MX93 or is it present in other CPU / MCU (we us a lot of different NXP products in our company)
Hello,
In Linux OS, since software latency is enough, this issue will not be observed, no need extra workaround.
In real time OS like QNX, extra delay time should be added to avoid this issue. So in u-boot, 10μs delay is added for uSDHC read transfer since Linux 6.6.36_2.1.0.
This errata will not affect application layer.
Best Regards,
Zhiming
Hello,
Is this commit?
Best Regards,
Zhiming