Hi folks,

I've been doing some work with very large MTD devices (many GB) and also
with improving YAFFS2 performance on those. There are some MTD patches
which will be forthcoming for 64-bit MTD (for which I needed to use a
few do_div operations as YAFFS2/MTD code can't use convenient shifts)
but in the meantime, here are some notes that I have on YAFFS2 -
these are from my experiences on a custom backport to
Montavista Linux Pro 3.1 (kernel 2.4.20).

None of this is meant to criticise - I think YAFFS2 is amasingly cool -
but here are some comments based on a particular effort.

* YAFFS2 should not use any shortopcaches when on Linux. It seems to
  make little difference to performance (only negative).

* YAFFS2 memory allocation using kmalloc does not work on very large
  devices and needs to use vmalloc instead in those cases (>2GB devices).
  The lack of checking for success proves to be a problem.

* YAFFS2 has various internal usage of types which makes it difficult to
  scale to >2GB devices. We have to divide up into multiple partitions.

* Andre Renaud latched onto a problem which I then rediscovered in
  performance testing. Having chunk groups of 16 reduces performance by
  at least 50% but in practice can be much higher. By applying a version
  of his patch, I was able ot reduce read time for a 50MB file from 27
  seconds to around 15 seconds and have achieved sustained reads at
  22.2Mbit/s on multi-GB devices reading many hundred MBs.

* YAFFS2 makes use of some additional reads and memcpy's which don't
  seem entirely necessary - by combining and changing some of the logic
  it looks like we could get another 10% performance gain.

Jon.