FreeBSD's embedded agenda

[Poul-Henning Kamp]

In message <4475EFC1.1020504 at nortel.com>, "Andrew Atrens" writes:

>>>Said kernel would have a low level driver that makes plain
>>>old flash chips look (and behave) like a disk. It would support
>>>wear-levelling, [...]
>>>
>>>Then you could throw FFS on top of that.
>> 
>> This is exactly what you do not want to do.
>> 
>> You want to write a flash friendly filesystem which knows what
>> a flash is, and which does the wear levelling internally.
>> 
>> The reason Flash Adaptation Layers came about in the first place
>> is that W95 didn't support anything but FAT.
>
>Hmm. I was thinking about partitioning the problem actually. Make flash
>look like a disk and then you can put any filesystem on it that you
>want. Seems a heck of a lot simpler .. and I'm not sure if I see any
>drawbacks to doing it that way ...

The main one is that the flash adaptation layer does not have the
full information to work with for deciding wear-leveling decisions
and the filesystem has no idea what the optimal block allocation
strategy is for the flash device.

Flash devices have no seek time penalty, and therefore the block
allocation should focus on wear-leveling rather than seek time
optimization.

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.