zephyr: NXP RT1061: First Byte Missing When Reading from W25Q512NW via FlexSPI2 at 30 MHz

Describe the bug

Hi,

I’m working with a custom board based on the NXP RT1061 SoC. We have enabled the FlexSPI2 peripheral and connected it to a Winbond W25Q512NW NOR flash device. Due to board constraints and signal integrity considerations, we are operating the bus at ~30 MHz instead of the typical 133 MHz.

Below is the relevant device tree configuration:

&flexspi2 {
    reg = <0x402a4000 0x4000>, <0x70000000 DT_SIZE_M(64)>;
    status = "okay";
    pinctrl-0 = <&pinmux_flexspi2>;
    pinctrl-names = "default";

    w25q512nw1: w25q512nw1@0 {
        compatible = "nxp,imx-flexspi-nor";
        size = <DT_SIZE_M(64 * 8)>;
        reg = <0>;
        spi-max-frequency = <DT_FREQ_M(30)>;
        status = "okay";
        jedec-id = [ef 60 20];
        erase-block-size = <4096>;
        write-block-size = <1>;
        cs-hold-time = <10>;
        cs-setup-time = <10>;

        partitions {
            compatible = "fixed-partitions";
            #address-cells = <1>;
            #size-cells = <1>;

            w25q512nw1_partition: partition@0 {
                label = "NOR1";
                reg = <0x00000000 DT_SIZE_M(64)>;
            };
        };
    };
};

Setting spi-max-frequency = <DT_FREQ_M(30)>; alone was not sufficient to enforce the desired frequency. I also had to manually configure the PLL and clock settings in the board initialization code:

CLOCK_DisableClock(kCLOCK_FlexSpi2);
CLOCK_SetDiv(kCLOCK_Flexspi2Div, 7);
CLOCK_SetMux(kCLOCK_Flexspi2Mux, 2);
CLOCK_EnableClock(kCLOCK_FlexSpi2);

The driver for this flash uses the Fast Read Quad I/O (0xEC) command with 8 dummy cycles.

Image

Problem:

When issuing read request, the first byte is consistently missing. The driver defaults to 8 dummy cycles, but according to the W25Q512NW datasheet (section 8.2.23 — Set Read Parameters), the device powers up with a default of 6 dummy cycles, and it specifies: Image This likely happens because the flash expects 6 dummy cycles, but the RT interprets the first valid data byte as the last dummy cycle. It’s unclear if 8 dummy cycles are valid at low frequencies when explicitly configured, but in practice, we observe data loss with the current driver setup.

Zephyr version 4.1.

Regression

  • This is a regression.

Steps to reproduce

No response

Relevant log output


Impact

Annoyance – Minor irritation; no significant impact on usability or functionality.

Environment

No response

Additional Context

No response

About this issue

  • Original URL
  • State: open
  • Created a year ago
  • Comments: 40 (40 by maintainers)

Most upvoted comments

How does your response answer my question? I was asking why the Zephyr test test_read_unaligned_address passes, but the synthetic test I created fails -despite both performing unaligned reads.

@ofirshe .Sorry for taking my assumption to your question. According to my testing, if you do not do unalignment write then it is OK for unalignment read even for your case, the default test_read_unaligned_address in zephyr does not do the unaligned write. But why the unalignment write fails, need more investigation, do you happen remember how do you resolve the cache issue? Thanks.

As I mentioned above, I’m just adjusting the dummy cycles - with 6 I get the expected result, but with 8 I consistently lose the first byte. I also see coherence issues in other tests I’ve run, but I haven’t yet been able to identify the root cause. Maybe you should consult with FlexSPI experts regarding the issue you encountered. Can you please open a ticket for that as well ?

Could you please try to reproduce the issue on zephyr/boards/nxp/mimxrt1170_evk/mimxrt1170_evk_mimxrt1176?

I need move the code to ram and try to use 30M timing. will feedback