LMK04828: LMK Programming failed in U-boot

Hi Team,

We are currently working on the LMK04828 clock synthesizer bring-up on an Intel Agilex-9 platform and would like your guidance regarding SPI access during the U-Boot stage.

Platform Overview

• SoC: Intel Agilex-9

• SPI controller: DesignWare SPI (dw_spi)

• Bootloader: U-Boot (DM_SPI framework)

• Kernel: Linux (SPI access working correctly)

• CS type: GPIO-based CS defined via Device Tree

• SPI mode: Mode 0

• SPI frequency: 1 MHz

Current Status (Important)

• LMK04828 SPI programming works correctly in the Linux kernel

•  SPI access to LMK04828 fails only in U-Boot

This confirms that:

• Hardware connections are correct

• SPI timing and protocol are correct

• LMK04828 device behavior is as expected

• The issue appears to be specific to the U-Boot SPI/CS handling, not the LMK04828 device itself

Observed Issue in U-Boot

While attempting to program LMK04828 in U-Boot using raw SPI write/read-back sequences, we encounter the following error:

Invalid chip select (err = -19 / ENODEV)

This error occurs during chip-select validation in the U-Boot SPI framework before actual data transfer.

Additional Observation

On the same SPI controller and U-Boot environment, when a QSPI flash device is connected:

• SPI access works correctly

• Chip select validation succeeds

• No CS-related errors are observed

This indicates that:

• The SPI controller and GPIO-based CS are functional in U-Boot

• The issue appears specific to non-flash SPI slave handling in U-Boot, such as LMK04828

SPI Access Method in U-Boot

• Using U-Boot DM_SPI APIs only

• CS is defined in the Device Tree using cs-gpios

• No manual CS GPIO handling in board code

• Raw SPI access (write + read-back), as required for LMK04828

• No legacy SPI APIs are used

Clarification Requested

We would appreciate TI’s guidance on the following:

1. Does TI recommend any specific SPI driver or access method for LMK04828 during the bootloader or bare-metal stages?

2. Is there any U-Boot–specific reference implementation for LMK04828 SPI programming?

3. Are there any special CS timing or sequencing requirements (e.g., CS setup/hold time) that bootloaders must explicitly handle?

4. Are there any known differences or constraints when accessing LMK04828 from a bootloader environment versus Linux?

5. Does TI have any known errata or recommendations related to LMK04828 SPI usage during early boot?

Summary

• LMK04828 works correctly in Linux

• SPI and CS hardware are validated

• The Issue is isolated to U-Boot SPI handling

• Guidance is requested specifically for bootloader-stage access

Please let us know if you need:

• SPI waveform captures

• Linux driver reference

• U-Boot source snippets

• Device Tree excerpts

Thank you for your support.

Best regards,
Thaniyel