NAND FLASH commands and notes

See NOTE below!!!

# (C) Copyright 2003
# Dave Ellis, SIXNET, dge@sixnetio.com
#
# SPDX-License-Identifier:	GPL-2.0+

Commands:

   nand bad
      Print a list of all of the bad blocks in the current device.

   nand device
      Print information about the current NAND device.

   nand device num
      Make device `num' the current device and print information about it.

   nand erase off|partition size
   nand erase clean [off|partition size]
      Erase `size' bytes starting at offset `off'. Alternatively partition
      name can be specified, in this case size will be eventually limited
      to not exceed partition size (this behaviour applies also to read
      and write commands). Only complete erase blocks can be erased.

      If `erase' is specified without an offset or size, the entire flash
      is erased. If `erase' is specified with partition but without an
      size, the entire partition is erased.

      If `clean' is specified, a JFFS2-style clean marker is written to
      each block after it is erased.

      This command will not erase blocks that are marked bad. There is
      a debug option in cmd_nand.c to allow bad blocks to be erased.
      Please read the warning there before using it, as blocks marked
      bad by the manufacturer must _NEVER_ be erased.

   nand info
      Print information about all of the NAND devices found.

   nand read addr ofs|partition size
      Read `size' bytes from `ofs' in NAND flash to `addr'.  Blocks that
      are marked bad are skipped.  If a page cannot be read because an
      uncorrectable data error is found, the command stops with an error.

   nand read.oob addr ofs|partition size
      Read `size' bytes from the out-of-band data area corresponding to
      `ofs' in NAND flash to `addr'. This is limited to the 16 bytes of
      data for one 512-byte page or 2 256-byte pages. There is no check
      for bad blocks or ECC errors.

   nand write addr ofs|partition size
      Write `size' bytes from `addr' to `ofs' in NAND flash.  Blocks that
      are marked bad are skipped.  If a page cannot be read because an
      uncorrectable data error is found, the command stops with an error.

      As JFFS2 skips blocks similarly, this allows writing a JFFS2 image,
      as long as the image is short enough to fit even after skipping the
      bad blocks.  Compact images, such as those produced by mkfs.jffs2
      should work well, but loading an image copied from another flash is
      going to be trouble if there are any bad blocks.

   nand write.trimffs addr ofs|partition size
      Enabled by the CONFIG_CMD_NAND_TRIMFFS macro. This command will write to
      the NAND flash in a manner identical to the 'nand write' command
      described above -- with the additional check that all pages at the end
      of eraseblocks which contain only 0xff data will not be written to the
      NAND flash. This behaviour is required when flashing UBI images
      containing UBIFS volumes as per the UBI FAQ[1].

      [1] http://www.linux-mtd.infradead.org/doc/ubi.html#L_flasher_algo

   nand write.oob addr ofs|partition size
      Write `size' bytes from `addr' to the out-of-band data area
      corresponding to `ofs' in NAND flash. This is limited to the 16 bytes
      of data for one 512-byte page or 2 256-byte pages. There is no check
      for bad blocks.

   nand read.raw addr ofs|partition [count]
   nand write.raw addr ofs|partition [count]
      Read or write one or more pages at "ofs" in NAND flash, from or to
      "addr" in memory.  This is a raw access, so ECC is avoided and the
      OOB area is transferred as well.  If count is absent, it is assumed
      to be one page.  As with .yaffs2 accesses, the data is formatted as
      a packed sequence of "data, oob, data, oob, ..." -- no alignment of
      individual pages is maintained.

Configuration Options:

   CONFIG_CMD_NAND
      Enables NAND support and commmands.

   CONFIG_CMD_NAND_TORTURE
      Enables the torture command (see description of this command below).

   CONFIG_MTD_NAND_ECC_JFFS2
      Define this if you want the Error Correction Code information in
      the out-of-band data to be formatted to match the JFFS2 file system.
      CONFIG_MTD_NAND_ECC_YAFFS would be another useful choice for
      someone to implement.

   CONFIG_SYS_MAX_NAND_DEVICE
      The maximum number of NAND devices you want to support.

   CONFIG_SYS_NAND_MAX_ECCPOS
      If specified, overrides the maximum number of ECC bytes
      supported.  Useful for reducing image size, especially with SPL.
      This must be at least 48 if nand_base.c is used.

   CONFIG_SYS_NAND_MAX_OOBFREE
      If specified, overrides the maximum number of free OOB regions
      supported.  Useful for reducing image size, especially with SPL.
      This must be at least 2 if nand_base.c is used.

   CONFIG_SYS_NAND_MAX_CHIPS
      The maximum number of NAND chips per device to be supported.

   CONFIG_SYS_NAND_SELF_INIT
      Traditionally, glue code in drivers/mtd/nand/nand.c has driven
      the initialization process -- it provides the mtd and nand
      structs, calls a board init function for a specific device,
      calls nand_scan(), and registers with mtd.

      This arrangement does not provide drivers with the flexibility to
      run code between nand_scan_ident() and nand_scan_tail(), or other
      deviations from the "normal" flow.

      If a board defines CONFIG_SYS_NAND_SELF_INIT, drivers/mtd/nand/nand.c
      will make one call to board_nand_init(), with no arguments.  That
      function is responsible for calling a driver init function for
      each NAND device on the board, that performs all initialization
      tasks except setting mtd->name, and registering with the rest of
      U-Boot.  Those last tasks are accomplished by calling  nand_register()
      on the new mtd device.

      Example of new init to be added to the end of an existing driver
      init:

	/*
	 * devnum is the device number to be used in nand commands
	 * and in mtd->name.  Must be less than
	 * CONFIG_SYS_NAND_MAX_DEVICE.
	 */
	mtd = &nand_info[devnum];

	/* chip is struct nand_chip, and is now provided by the driver. */
	mtd->priv = &chip;

	/*
	 * Fill in appropriate values if this driver uses these fields,
	 * or uses the standard read_byte/write_buf/etc. functions from
	 * nand_base.c that use these fields.
	 */
	chip.IO_ADDR_R = ...;
	chip.IO_ADDR_W = ...;

	if (nand_scan_ident(mtd, CONFIG_SYS_MAX_NAND_CHIPS, NULL))
		error out

	/*
	 * Insert here any code you wish to run after the chip has been
	 * identified, but before any other I/O is done.
	 */

	if (nand_scan_tail(mtd))
		error out

	if (nand_register(devnum))
		error out

      In addition to providing more flexibility to the driver, it reduces
      the difference between a U-Boot driver and its Linux counterpart.
      nand_init() is now reduced to calling board_nand_init() once, and
      printing a size summary.  This should also make it easier to
      transition to delayed NAND initialization.

      Please convert your driver even if you don't need the extra
      flexibility, so that one day we can eliminate the old mechanism.


   CONFIG_SYS_NAND_ONFI_DETECTION
	Enables detection of ONFI compliant devices during probe.
	And fetching device parameters flashed on device, by parsing
	ONFI parameter page.

   CONFIG_BCH
	Enables software based BCH ECC algorithm present in lib/bch.c
	This is used by SoC platforms which do not have built-in ELM
	hardware engine required for BCH ECC correction.

   CONFIG_SYS_NAND_BUSWIDTH_16BIT
	Indicates that NAND device has 16-bit wide data-bus. In absence of this
	config, bus-width of NAND device is assumed to be either 8-bit and later
	determined by reading ONFI params.
	Above config is useful when NAND device's bus-width information cannot
	be determined from on-chip ONFI params, like in following scenarios:
	- SPL boot does not support reading of ONFI parameters. This is done to
	  keep SPL code foot-print small.
	- In current U-Boot flow using nand_init(), driver initialization
	  happens in board_nand_init() which is called before any device probe
	  (nand_scan_ident + nand_scan_tail), thus device's ONFI parameters are
	  not available while configuring controller. So a static CONFIG_NAND_xx
	  is needed to know the device's bus-width in advance.
	Some drivers using above config are:
	drivers/mtd/nand/mxc_nand.c
	drivers/mtd/nand/ndfc.c
	drivers/mtd/nand/omap_gpmc.c


Platform specific options
=========================
   CONFIG_NAND_OMAP_GPMC
	Enables omap_gpmc.c driver for OMAPx and AMxxxx platforms.
	GPMC controller is used for parallel NAND flash devices, and can
	do ECC calculation (not ECC error detection) for HAM1, BCH4, BCH8
	and BCH16 ECC algorithms.

   CONFIG_NAND_OMAP_ELM
	Enables omap_elm.c driver for OMAPx and AMxxxx platforms.
	ELM controller is used for ECC error detection (not ECC calculation)
	of BCH4, BCH8 and BCH16 ECC algorithms.
	Some legacy platforms like OMAP3xx do not have in-built ELM h/w engine,
	thus such SoC platforms need to depend on software library for ECC error
	detection. However ECC calculation on such plaforms would still be
	done by GPMC controller.

   CONFIG_NAND_OMAP_ECCSCHEME
	On OMAP platforms, this CONFIG specifies NAND ECC scheme.
	It can take following values:
	OMAP_ECC_HAM1_CODE_SW
		1-bit Hamming code using software lib.
		(for legacy devices only)
	OMAP_ECC_HAM1_CODE_HW
		1-bit Hamming code using GPMC hardware.
		(for legacy devices only)
	OMAP_ECC_BCH4_CODE_HW_DETECTION_SW
		4-bit BCH code (unsupported)
	OMAP_ECC_BCH4_CODE_HW
		4-bit BCH code (unsupported)
	OMAP_ECC_BCH8_CODE_HW_DETECTION_SW
		8-bit BCH code with
		- ecc calculation using GPMC hardware engine,
		- error detection using software library.
		- requires CONFIG_BCH to enable software BCH library
		(For legacy device which do not have ELM h/w engine)
	OMAP_ECC_BCH8_CODE_HW
		8-bit BCH code with
		- ecc calculation using GPMC hardware engine,
		- error detection using ELM hardware engine.
	OMAP_ECC_BCH16_CODE_HW
		16-bit BCH code with
		- ecc calculation using GPMC hardware engine,
		- error detection using ELM hardware engine.

	How to select ECC scheme on OMAP and AMxx platforms ?
	-----------------------------------------------------
	Though higher ECC schemes have more capability to detect and correct
	bit-flips, but still selection of ECC scheme is dependent on following
	- hardware engines present in SoC.
		Some legacy OMAP SoC do not have ELM h/w engine thus such
		SoC cannot support BCHx_HW ECC schemes.
	- size of OOB/Spare region
		With higher ECC schemes, more OOB/Spare area is required to
		store ECC. So choice of ECC scheme is limited by NAND oobsize.

	In general following expression can help:
		NAND_OOBSIZE >= 2 + (NAND_PAGESIZE / 512) * ECC_BYTES
	where
		NAND_OOBSIZE	= number of bytes available in
				OOB/spare area per NAND page.
		NAND_PAGESIZE	= bytes in main-area of NAND page.
		ECC_BYTES	= number of ECC bytes generated to
				protect 512 bytes of data, which is:
				3 for HAM1_xx ecc schemes
				7 for BCH4_xx ecc schemes
				14 for BCH8_xx ecc schemes
				26 for BCH16_xx ecc schemes

		example to check for BCH16 on 2K page NAND
		NAND_PAGESIZE = 2048
		NAND_OOBSIZE = 64
		2 + (2048 / 512) * 26 = 106 > NAND_OOBSIZE
		Thus BCH16 cannot be supported on 2K page NAND.

		However, for 4K pagesize NAND
		NAND_PAGESIZE = 4096
		NAND_OOBSIZE = 64
		ECC_BYTES = 26
		2 + (4096 / 512) * 26 = 210 < NAND_OOBSIZE
		Thus BCH16 can be supported on 4K page NAND.


NOTE:
=====

The current NAND implementation is based on what is in recent
Linux kernels.  The old legacy implementation has been removed.

If you have board code which used CONFIG_NAND_LEGACY, you'll need
to convert to the current NAND interface for it to continue to work.

The Disk On Chip driver is currently broken and has been for some time.
There is a driver in drivers/mtd/nand, taken from Linux, that works with
the current NAND system but has not yet been adapted to the u-boot
environment.

Additional improvements to the NAND subsystem by Guido Classen, 10-10-2006

JFFS2 related commands:

  implement "nand erase clean" and old "nand erase"
  using both the new code which is able to skip bad blocks
  "nand erase clean" additionally writes JFFS2-cleanmarkers in the oob.

Miscellaneous and testing commands:
  "markbad [offset]"
  create an artificial bad block (for testing bad block handling)

  "scrub [offset length]"
  like "erase" but don't skip bad block. Instead erase them.
  DANGEROUS!!! Factory set bad blocks will be lost. Use only
  to remove artificial bad blocks created with the "markbad" command.

  "torture offset"
  Torture block to determine if it is still reliable.
  Enabled by the CONFIG_CMD_NAND_TORTURE configuration option.
  This command returns 0 if the block is still reliable, else 1.
  If the block is detected as unreliable, it is up to the user to decide to
  mark this block as bad.
  The analyzed block is put through 3 erase / write cycles (or less if the block
  is detected as unreliable earlier).
  This command can be used in scripts, e.g. together with the markbad command to
  automate retries and handling of possibly newly detected bad blocks if the
  nand write command fails.
  It can also be used manually by users having seen some NAND errors in logs to
  search the root cause of these errors.
  The underlying nand_torture() function is also useful for code willing to
  automate actions following a nand->write() error. This would e.g. be required
  in order to program or update safely firmware to NAND, especially for the UBI
  part of such firmware.


NAND locking command (for chips with active LOCKPRE pin)

  "nand lock"
  set NAND chip to lock state (all pages locked)

  "nand lock tight"
  set NAND chip to lock tight state (software can't change locking anymore)

  "nand lock status"
  displays current locking status of all pages

  "nand unlock [offset] [size]"
  unlock consecutive area (can be called multiple times for different areas)

  "nand unlock.allexcept [offset] [size]"
  unlock all except specified consecutive area

I have tested the code with board containing 128MiB NAND large page chips
and 32MiB small page chips.