search

littlefs-project / littlefs

A little fail-safe filesystem designed for microcontrollers
BSD 3-Clause "New" or "Revised" License
5.12k stars 790 forks source link

user_provided_block_device_prog #932

Closed JohannesLichtenberger closed 7 months ago

JohannesLichtenberger commented 8 months ago

Does this function have to write a single page/block or does it have to account for writing multiple pages? I think, regarding the source only a single block gets flushed everytime, so that might be sufficient already!? :-)

kind regards Johannes

geky commented 8 months ago

Hi @JohannesLichtenberger, the bd's read and prog functions don't cross block_size boundaries, but they can be called with requests that are multiples of the read_size and prog_size.

prog_size is sometimes equivalent to "page" on some devices, sometimes not. Unfortunately the terminology gets a bit confusing and conflicting when you look at storage across different technologies...

JohannesLichtenberger commented 8 months ago

Oh ok, so I'll have to use a function, which can write several pages to a Winbond W25Q :-)

geky commented 8 months ago

The benefit of this approach is sometimes hardware can support faster writes over larger prog sizes. This wouldn't be possible if littlefs broke up the progs for you.

Winbond W25Q

I believe this chip supports byte-level progs. You could set prog_size=1 and then opportunistically use larger prog operations in your prog function. This would reduce the overhead of metadata padding.

But that depends on how much effort you want to put in. There's nothing functionally wrong with programming a full page every time.