Wear leveling issue

[arc3]

Hello, there.
I haven't found information on this issue, so I'm asking here.
Does wear leveling feature work correctly with constant files?
I'm testing spiffs and I've come across next situation:
presets:
phys_size = 1Mb
phys_sector_size = 64Kb
log_block_size = 64Kb
log_page_size = 1024b

1. I write some big file. Size of 500Kb.
2. I start multiple writing&removing operations for smaller file. Size of 50Kb.
3. Then I check block erases and see next:
   -------------------after 499 rewriting small file
   0000 ddddd//ddddddddddddddddd/////////////////////////////////////// era_cnt: 555
   0001 ddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddd era_cnt: 0
   0002 ddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddd era_cnt: 0
   0003 ddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddd era_cnt: 0
   0004 ddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddd era_cnt: 0
   0005 ddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddd era_cnt: 0
   0006 ddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddddd era_cnt: 0
   0007 _______________________________________________________________ era_cnt: 561
   0008 _______________________________________________________________ era_cnt: 562
   0009 //////////////////////////ddddddddddddd//////////////////////// era_cnt: 558
   000a /////////////////////////////ddddddddddddddddd///////////////// era_cnt: 560
   000b //dddddddddddddddddddddddddddddddddddd/ddddd/////////////////// era_cnt: 551
   000c ////////////////////////////////////dddddddddddddddd/////////// era_cnt: 559
   000d //////////////////////////////////////////dddddddddidddddddiidd era_cnt: 557
   000e ddddddddddddddddddddddddddddddddddddddddddddddddd______________ era_cnt: 556
   000f _______________________________________________________________ era_cnt: 563

Those "era_cnt: 0" means that fully written blocks do not participate in wear leveling.
Am I right?
I've played with spiffs_gc_find_candidate() here
https://github.com/pellepl/spiffs/blob/master/src/spiffs_gc.c#L297
Removed "deleted_pages_in_block > 0" condition and GC began to include fully written blocks
into candidates list. Because of erase age part of score calculation.
After that erases became more or less the same for all blocks.
I had to increase SPIFFS_GC_MAX_RUNS also.
But, of course, total number of read&writes increased too.
So, the question is:
Is the behaviour from my example correct and expected?
p.s. "Static wear leveling works the same as dynamic wear leveling except the static blocks that do not change are periodically moved so that these low usage cells are able to be used by other data." by wikipedia :-)
p.p.s. sorry for my bad English. ))

[pellepl]

Hi,

all your assumptions are correct. That's a bug you found there! Spiffs is supposed to recycle low usage cells also.

Luckily, this case will probably not happen very often - actually filling a block with data calls for having big logical page size and being able to write a full logical block of data in one go (which seldom is the case in embedded business).

What I have to figure out what the reason was for adding deleted_pages_in_block > 0. Now, four years later, it seems stupid.

Thanks for finding this!

[pellepl]

Well, I cannot find a single reason to why deleted_pages_in_block > 0 was added. Since all tests pass when removing it, I'll do so.

Cheers / P