am335x_spl_bch - SPL NAND loading hang on power on only

Kevin Baker kbaker at gmail.com
Tue Aug 31 02:55:16 CEST 2021


 Hi all,

I am working on updating our version of u-boot, and came across something
weird - the SPL hangs after "Trying to load NAND" on the first power-on
boot, but after a watchdog reset or power-on reset, the NAND works fine.

    U-Boot SPL 2021.01-00010-g8c91cf967b-dirty (Aug 30 2021 - 19:03:45
-0500)
    Trying to boot from NAND        <<<<   hangs here until watchdog reset

After checking more common issues - pinmux/config/timing/NAND connections,
etc. and printf-tracing it, it is actually hanging after board_nand_init
and at this call to select_chip before getting to NAND_CMD_RESET:

drivers/mtd/nand/raw/am335x_spl_bch.c: nand_init()
...
    board_nand_init(&nand_chip);

    if (nand_chip.select_chip)
        nand_chip.select_chip(mtd, 0);    <<<< stuck here ??? <<<<

    /* NAND chip may require reset after power-on */
    nand_command(0, 0, 0, NAND_CMD_RESET);
...

Somehow, the first power up, nand_chip.select_chip is set to 0xFFFFFFFF, so
it hangs at that point. Whereas future power-on resets have the location
cleared to 0, so it works fine...

Explicitly setting select_chip to NULL in the call to board_nand_init for
omap_gpmc makes everything work fine:

drivers/mtd/nand/raw/omap_gpmc.c: board_nand_init()
...
    writel(gpmc_config, &gpmc_cfg->config)

+    nand->select_chip = NULL;
+
    nand->IO_ADDR_R = (void __iomem *)&gpmc_cfg->cs[cs].nand_dat;
...

But now I am worried about other things not being cleared out properly at
reset... The am335x SPL NAND driver is pretty old and presumably
well-tested code, so I was kinda surprised at this fix working. Would
anyone know if a configuration setting, compiler setting or similar
contributes to this? Or is it some artifact of the power-up state of the
board?

Is there a better way of 'clearing' the static nand_chip struct in
am335x_spl_bch.c, or is there really a missing setup of this function
pointer somewhere for am335x_spl_bch/omap_gpmc?

This is against TI's u-boot @ d77252021a85c72d7ff9a4e151b81c4b4b9e1205 ,
with a Variscite VAR-SOM-AM33 module. Just wanted to check with the list to
see if I am missing something...

Thanks,
Kevin


More information about the U-Boot mailing list