On Wednesday 07 March 2007 15:37, Brad Beveridge wrote:
> On 06/03/07, Charles Manning <manningc2@actrix.gen.nz> wrote:
> > On Wednesday 07 March 2007 15:06, Johnson, Charles F wrote:
> > > Can YAFFS2 support the following:
> > >
> > > * Virtual page: 4KB, 8KB or 16KB
>
> I'm sure you're aware, but one downside of large virtual blocks is
> that you will have much greater wastage when a block is marked bad.
It comes down to how you structure the blocks.
If you use 16kB virtual pages (yaffs term is "chunks") with 64 chunks per
block, this might be achieved by running 8x2kbyte style devices in parallel.
That would certainly raise the issue that Brad mentions since if any block is
bad, it will take out of service all the parallel blocks in the "virtual
block".
You can also run fewer chunks per block. For example, you could still use
128kbyte blocks (native size of 2k page devices) by setting this up as 16
chunks of 8kB instead of the more conventional 64 chunks of 2kB. This reduces
the tnode overhead considerably without causing the parallel block loss Brad
raises.
The chunks per block number is very flexible. I have not tried very low
numbers (1, 2 etc) but there is no reason to think that won't work. People
using larger numbers (eg. 248 in one case where someone is running YAFFS on
NOR).
-- Charles
(text/plain)