[chanijjani]

• The freecount value of an erasable block mismatches with
  actual block status when MTD_ERASE fails.
  As a result, no physical sector is allocated and file system errors
  occur at various moments.
  Error handling code is added to avoid this situation.
• Useful debug messages are added also.

[seinfra]

Target : [c8dec000062c5b2eabf9d6aa782edee820d0fd5f] - Code Rule Check OK.

[seinfra]

Target : [c8dec000062c5b2eabf9d6aa782edee820d0fd5f] - Code Rule Check (C++) OK.

[Taejun-Kwon]

IMO, you have to clean up this & current code
As you know, flow is below.

1. Check all erase block and check count of free-sector of each block.
2. If there's free sector in specific block, check wear level.
   2-1. if wearlevel is acceptable, checked block becomes a target block
3. If there's no free sector after flow Merge Posix Cancellation Point #2, relocate block & retry Create README.md #1~Merge Posix Cancellation Point #2
4. Now we found target block, then check all sector header in block.
5. If it found sector which has clean header, then verify that contents area's data.
   5-1 if any of contents is flipped, update erase count and something, then retry it
6. If any of sector is all cleaned, then allocate it, otherwise it is program bug.

So your approach is to prevent #6 because of invalid header, I think, am I right?
Then this is not efficient code, because we already check header's status in #6.
I think you can modify something in below.

		if ((UINT8TOUINT16(header.logicalsector) == 0xFFFF) &&
#if SMART_STATUS_VERSION == 1
			((header.seq == 0xFF) && (header.crc8 == 0xFF)) &&
#else
			(header.seq == CONFIG_SMARTFS_ERASEDSTATE) &&
#endif
			(!(SECTOR_IS_COMMITTED(header)))) {

This can be modify to HEADER_IS_CLEAN but HEADER_IS_CLEAN doesn't care CONFIG_SMARTFS_ERASEDSTATE, you have to modify it too.

if (HEADER_IS_CLEAN(header)) {
    check_bitflip_in_contents;
} else {
    compare_with_sMap;
    if (sMap is valid but header is not clean) {
          release_sector;
          update_freecount & release_count;
   }
   ...
}

[Taejun-Kwon]

Also if allocation falied because of flipped bit, retry is necessary for both of if-else in my comment.

[chanijjani]

@Taejun-Kwon Thanks for your well-organized summary.

I uploaded this PR since a situation that the 'Program bug!' message in line 3605 happened during easy-setup tests on NXP boards.

This is due to the phenomenon that the freecount value of an erasable block(=EB) selected in line 3433 has no clean sectors in actual so that no physical sectors in the EB passed "if" statement in line 3563. As a result, the execution reaches at 3605 since no physical sector is allocated even though an EB is chosen.

Now I'm trying to find the root cause and fix it in a more efficient way.

[seinfra]

Target : [ec87addbe79a2ff1b488ae11d608b72da1779021] - Code Rule Check (C++) OK.

[seinfra]

Target : [ec87addbe79a2ff1b488ae11d608b72da1779021] - Code Rule Check OK.

[seinfra]

Target : [985761b45cec8a8503e877716a851672f69d6e60] - Code Rule Check (C++) OK.

[seinfra]

Target : [985761b45cec8a8503e877716a851672f69d6e60] - Code Rule Check OK.

[seinfra]

Target : [04cbef6d70776f48158bdbe599bf296da3aea7d6] - Code Rule Check (C++) OK.

[seinfra]

Target : [04cbef6d70776f48158bdbe599bf296da3aea7d6] - Code Rule Check OK.

[seinfra]

Target : [cde39e5b96118685e61c02844c41f99bd653165d] - Code Rule Check (C++) OK.

[seinfra]

Target : [cde39e5b96118685e61c02844c41f99bd653165d] - Code Rule Check OK.

[seinfra]

Target : [bf4904f] - Code Rule Check (C++) OK.

[seinfra]

Target : [bf4904f] - Code Rule Check OK.