Scribd
Upload
118 views7 pages

NAND Flash Bad Block Management

Uploaded by

m3y54m
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
118 views7 pages

NAND Flash Bad Block Management

Uploaded by

m3y54m
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

Application Note

freescale
semiconductor
TM
08 10, 2010

NAND Flash Bad Block Management


---For Linux BSP

ABSTRACT:

This doc describe how to handle nand flash bad block with linux software
KEYWORDS: NAND BAD BLOCK BBI BBT

APPROVED:

AUTHOR COMMENTS

Jason Liu Initial draft

FREESCALE SEMICONDUCTOR INTERNAL USE ONLY

© FREESCALE SEMICONDUCTOR, INC. 2010


All Rights Reserved. Presence of copyright notice is not an acknowledgement of publication
Application Note

NAND Flash Bad Block Management


---For Linux BSP

1. What is bad block?


Nand flash will have some invalid blocks which is what we called bad blocks. This
invalid bad block can’t be used to store data because it’s not stable. Software should
avoid using these invalid blocks by means of checking the bad block first. If it’s bad, skip
it and not use it.

2. How to check bad block?


Nand flash manufactory will mark the bad block with one flag in the spare area of nand
flash out of factory. This bad block is what we called initial bad block. This flag is what
we called bad block indication (BBI). The location of BBI is defined by the NAND flash
manufacturer. Usually, the BBI will locate in the spare area as the following table shown.

Take 8bit NAND as example:

Nand Flash SLC(small page) SLC(large page) MLC


BBI offset in page 5th byte 1 1th byte 1 1th byte 1
Page offset in block first page first page First or first 2 pages or
Last page or last 2
pages 2
Note1: 0xFF means good block, others means bad block
Note2: The page which contains the bad block marker differs between each vendors. Refer to the NAND spec for
detailed information about which pages contain the bad block marker.

3. Incompatibility
Nand flash has the initial data lout which is main + spare area when it come out of factory.
Take 2KB + 64B MLC nand flash as example, the layout of nand flash is:

2112B total

2KB Main area 64B Spare

FREESCALE SEMICONDUCTOR INTERNAL USE ONLY

© FREESCALE SEMICONDUCTOR, INC. 2010


All Rights Reserved. Presence of copyright notice is not an acknowledgement of publication
Application Note

But the FSL IMX NFC layout is as the followings,

512B main 16B 512B main 16B 512B main 16B 512B main 16B

Section 1 Section 2 Section 3 Section 4

So, the BBI of NAND flash out of factory is located in the main area of sections 4, as the
following shows,

bad block indicator at column address 2048,


which is NAND out of factory layout

2KB Main area 64B Spare


spare

512B main 16B 512B main 16B 512B main 16B 512B 16B
spare spare spare main spare

Section 1 Section 2 Section 3 Section 4

0 527 0 527 0 527 0 527

bad block indicator at 4th Section, address offset is 464

Here is the summery of the incompatibility between the NAND flash layout and the FSL
NFC data layout:

 BBI of NAND flash out of factory is located in the data area of the last section of
NFC
 BBI byte of NFC layout is not correct with large page NAND flash, it only
compatible with small page NAND flash

In order to solve the incompatibility above, we need take one solution to preserve the
BBI out of factory not be overriden by user data, the solution is called BBI swap.

FREESCALE SEMICONDUCTOR INTERNAL USE ONLY

© FREESCALE SEMICONDUCTOR, INC. 2010


All Rights Reserved. Presence of copyright notice is not an acknowledgement of publication
Application Note

4. BBI swap solution internals


From the above graph, the user data will overwrite with BBI out of factory during the use
of NAND flash and make the BBI lost. So, we need figure out one solution to record the
BBI of factory when the NAND flash is used for the first time with FSL NAND flash
driver. The solution is as followings:

Solution: Take 2K + 64B MLC NAND as example


1) Swap the byte in main area which denotes the BBI with one byte in the spare area
during program, and let the BBI byte is 0xFF. The user data is now swapped to
the spare area.
2) Swap back the user data with the spare area which is used to store the user data
during read.
3) the software always check the swapped byte of spare area to get the BBI
information and build the bad block table for use.
bad block at column address 2048

2KB Main area 64B Spare

“real” “free”
Page in use

512B main 16B 512B main 16B 512B main 16B 512B 16B
spare spare spare main spare
Section 1 Section 2 Section 3 Section 4

0 527 0 527 0 527 0 527


For write: Move “real” data to spare area and put 0xFF
1: Before write, copy the 464th byte data into offset 464
in the main area RAM buffer to a “free”
spare-area RAM buffer. For read:
2. Write 0xFF to the 464th byte in the After reading data out of NAND flash,
main area RAM. copy the data from the “free” byte in the
3. Start “write” operation. spare-area RAM to the 464th byte in the
main area RAM buffer to “recover”

For bad block detection: Free byte offset:


After reading data out of NAND The second byte of last spare
flash, check the “free” byte section

FREESCALE SEMICONDUCTOR INTERNAL USE ONLY

© FREESCALE SEMICONDUCTOR, INC. 2010


All Rights Reserved. Presence of copyright notice is not an acknowledgement of publication
Application Note

5. Why Linux BSP need BBT solution?


BBT means bad block table which will be stored onto NAND flash.Without BBT, NAND
driver will scan all the blocks on NAND flash to get the bad block information . But with
BBT, NAND driver can just fetch the bad block information from bad blcok table which
has been stored onto NAND. We need BBT solution due to the following reasons:

 If run-time error occurs during write/erase, how to mark that block as bad
again? We may can't write to that block now
 Performance – without BBTgiven large size of NAND, the scanning time will
become pretty significant.

6. Why we have BBT solution, We still Need BBI swap?


Preserve the bad block indicator for NAND out of factory by using BBI swap will make
it possible for NAND driver to reconstruct the accurate BBT talbe once the bad block
table is been erased by other tools or bad block table gone corrupt or the bad blcok table
not exist.

7. NAND bad block management of Linux BSP


The following solution is the NAND bad block management applied into the Linux BSP:

Solution:(BBI swap + BBT)


1) Scan all the blocks to get the initial bad block information by checking the BBI
for a fresh NAND. BBI swap will be used to get the real BBI information from
the NAND out of factory.
2) Store the bad block information into the NAND flash . Thus all the bad block
information can be fetched from NAND during the following use. The bad block
information store on the NAND is what we called the BBT(Bad Block Table).
3) When run-time bad block happen, update BBT on the NAND flash.

FREESCALE SEMICONDUCTOR INTERNAL USE ONLY

© FREESCALE SEMICONDUCTOR, INC. 2010


All Rights Reserved. Presence of copyright notice is not an acknowledgement of publication
Application Note

Page in use
512B main 16B 512B main 16B 512B main 16B 512B 16B
spare spare spare main spare

Bad Block Table Store in one good block of the last 4 blocks.
There is one main table and one mirror table.
-3 -2 -1 -0

…….
Mirror Main

First Time Use: Follow up Use:


3) Scan all the blocks to get the initial 1) Read the BBT from the nand flash
bad block information by checking for the coming use of nand flash.
the BBI 2) When run-time bad block happen,
4) Store the bad block information, we update BBT on the nand flash.
also call it the bad block table
(BBT) into the nand flash.
.
For bad block mark:
For bad block detection: Update BBT on nand flash
Get the BBI from the BBT if BBT When block gone bad during
exist runtime.

8. NAND driver upgrade notes


We add the BBI swap solution in the Linux BSP 10.05 release which will cause some
incompatible issue when we try to upgrade Linux BSP release to 10.05 or higher version.

Issues:

The NAND data which written by old NAND driver will become corrupt when using the
new NAND driver. This is due to that one byte of user data written by old driver is
swapped to spare area and is replaced with 0xFF by new driver.

What need to do:

If the NAND flash contains the valid user data, please do the following,

1. Use old NAND driver to back up all the user files.


2. Use old NAND driver to erase all the NAND partitions
3. Use new NAND driver to copy back all the user files.

Note: If don’t want to do back/restore, just take step 2

As for the BBT table which created by old NAND driver. Currently, the BBT can’t be
erased by NAND driver, which means the BBT table, will remain there. The new driver
FREESCALE SEMICONDUCTOR INTERNAL USE ONLY

© FREESCALE SEMICONDUCTOR, INC. 2010


All Rights Reserved. Presence of copyright notice is not an acknowledgement of publication
Application Note

can recognize the BBT table and re-use it. But due to the BI swap, the bad block
information appears on the BBI offset will be swapped with 0xFF, which lead to any bad
block appears on this range will be taken as good block. This is very very rarely case and
if it’s unfortunately happen, the only we can do is to erase the bad block table with uboot
force erase function.

Recommendation:

Highly recommend using the BBI swap + BBT solutions for NAND flash bad block
management during mass production.

9.NAND driver internals


Please refer to drivers/mtd/nand/mxc_nd2.c & drivers/mtd/nand/mxc_nd2.h file under
Linux BSP source code package.

FREESCALE SEMICONDUCTOR INTERNAL USE ONLY

© FREESCALE SEMICONDUCTOR, INC. 2010


All Rights Reserved. Presence of copyright notice is not an acknowledgement of publication

You might also like