[U-Boot] Problem with attaching UBI partition

[Marek Vasut]

On 02/29/2016 05:44 PM, Bakhvalov, Denis (Nokia - PL/Wroclaw) wrote:
> Hi Jagan, Heiko,
> 
>> Did you enable CONFIG_SPI_FLASH_SPANSION on your config, becuase
>> S25FL512S is already been added u-boot. Pls- check the same and let me
>> know if you find any issues while detecting the flash.
> 
>> Whether the driver detecting flash or not with RDID, please define
>> DEBUG on drivers/mtd/spi/spi_flash.c
> 
> I had no problems with detecting the flash. It was detected like this:
> SF: Detected S25FL512S_256K with page size 512 Bytes, erase size 256 KiB, total 64 MiB
> 
> Yes, I have CONFIG_SPI_FLASH_SPANSION enabled in configs/socfpga_arria5_dbrrh_defconfig (dbrrh - is my custom board).
> 
> Actually, today I solved the problem, although in not very clean way, I think.
> 
> As I mentioned in previous mail I had problems with reading from the flash. Some bytes were read incorrectly.
> I'm migrating from U-Boot 2013 to mainline U-Boot. And on previous version there were no such issue.
> I found out that there was some workaround introduced in order to fix this.
> 
> So, I just apply some part of that workaround to mainline U-Boot and now it works.
> 
> Here is the problem in more detail:
> 
> On previous version (2013):
> U-BOOT # sf read 0x1B000000 0x03800000 4
>      cadence_qspi_apb_indirect_read_setup. addr_value = 0x3800000
>      cadence_qspi_apb_indirect_read_setup. rd_reg = 0x800000b
>      cadence_qspi_apb_indirect_read_setup. device size = 0x101003
> 
> U-BOOT # md 0x1B000000 1
> 1b000000: 55424923
> 
> But on mainline U-Boot(2016.03-rc1) I had:
> U-BOOT # sf read 0x1B000000 0x03800000 4
>      cadence_qspi_apb_indirect_read_setup. addr_value = 0x800000
>      cadence_qspi_apb_indirect_read_setup. rd_reg = 0x0b
>      cadence_qspi_apb_indirect_read_setup. device size = 0x101002
> 
> U-BOOT # md 0x1B000000 1
> 1b000000: 554249ff
> 
> You can see the difference in the register values that were written in QSPI registers.
> In case with mainline U-Boot I had address value not properly set.
> 
> Here is the workaround that I've made:
> 
> Index: drivers/mtd/spi/spi_flash.c
> ===================================================================
> --- drivers/mtd/spi/spi_flash.c	(revision 608)
> +++ drivers/mtd/spi/spi_flash.c	(revision 609)
> @@ -29,6 +29,17 @@
>  	cmd[3] = addr >> 0;
>  }
>  
> +#ifdef CONFIG_DBRRH_WORKAROUND                
> +static void spi_flash_addr32(u32 addr, u8 *cmd)
> +{
> +	/* cmd[0] is actual command */
> +	cmd[1] = (addr >> 24) & 0xFF;
> +	cmd[2] = (addr >> 16) & 0xFF;
> +	cmd[3] = (addr >> 8) & 0xFF;
> +	cmd[4] = (addr >> 0) & 0xFF;
> +}
> +#endif
> +
>  static int read_sr(struct spi_flash *flash, u8 *rs)
>  {
>  	int ret;
> @@ -510,8 +521,14 @@
>  		else
>  			read_len = remain_len;
>  
> +#ifdef CONFIG_DBRRH_WORKAROUND                
> +        		spi_flash_addr32(read_addr, cmd);
> +#else
>  		spi_flash_addr(read_addr, cmd);
> +#endif
>  
> 		ret = spi_flash_read_common(flash, cmd, cmdsz, data, read_len);
>  		if (ret < 0) {
>  			debug("SF: read failed\n");
> 
> Index: drivers/spi/cadence_qspi_apb.c
> ===================================================================
> --- drivers/spi/cadence_qspi_apb.c	(revision 608)
> +++ drivers/spi/cadence_qspi_apb.c	(revision 609)
> @@ -30,6 +30,10 @@
>  #include <asm/errno.h>
>  #include "cadence_qspi.h"
>  
> +#ifdef CONFIG_DBRRH_WORKAROUND
> +	#define CMD_OPTION_DUMMY_CYCLES		0x7F	/* Dummy Cycles for Read Command */
> +#endif
> +
> @@ -706,9 +710,25 @@
>  #endif
>  
>  	/* Get address */
> +#ifdef CONFIG_DBRRH_WORKAROUND
> +	addr_value = cadence_qspi_apb_cmd2addr(&cmdbuf[1], 4);
> +#else
>  	addr_value = cadence_qspi_apb_cmd2addr(&cmdbuf[1], addr_bytes);
> +#endif
> 
> +#ifdef CONFIG_DBRRH_WORKAROUND
> +	/* Setting Dummy Clock Cycle */
> +	dummy_clk = (0x08 & CMD_OPTION_DUMMY_CYCLES);
> +	if (dummy_clk) 
> +	{
> +		rd_reg |= (dummy_clk & CQSPI_REG_RD_INSTR_DUMMY_MASK)
> +			<< CQSPI_REG_RD_INSTR_DUMMY_LSB;
> +	}
> +#else
> +
>  	/* The remaining lenght is dummy bytes. */
>  	dummy_bytes = cmdlen - addr_bytes - 1;
>  	if (dummy_bytes) {
> @@ -731,7 +751,9 @@
>  			rd_reg |= (dummy_clk & CQSPI_REG_RD_INSTR_DUMMY_MASK)
>  				<< CQSPI_REG_RD_INSTR_DUMMY_LSB;
>  	}
> +#endif
> 
> 
> With this correction I can read contents of the flash properly.
> However, I'm a bit surprised that I was forced to make such correction like storing 4 bytes of address (see spi_flash_addr32() above).
> On the other hand I haven't found any switch that could be turned on to fix my problem in a clean and nice way.
> 
> With 24 bytes we can address only 16 MB. How cadence driver is supposed to work for larger spaces?
> Is this 4th byte comes from somewhere else?
> 
> Jagan, Heiko, please evaluate my correction.

Try enabling CONFIG_SPI_FLASH_BAR in your board config ;-)

I normally use the cadence QSPI driver with N25Q00AA, which is 128MiB
part. I also had the 4-byte addressing support in the works for a while,
but never got around to cleaning that up. It'd be real nice if you could
send a proper patch.

-- 
Best regards,
Marek Vasut