Re: Pooled memory allocation for JFFS2

[David Woodhouse <dwmw2 at infradead dot org>]

On Thu, 2003-11-20 at 09:54 -0700, Gary Thomas wrote:
> Committed to main CVS, thanks, David :-)

Thanks.

Note this includes the optional compression, and also makes the
debuglevel a CDL option -- but it's lacking Thomas' pooled memory
allocation.

Upon implementing ilookup() to match the current Linux kernels, I
observe that the inode use count handling in iget() and iput() of
fs-ecos.c is still very dubious.

iget() and ilookup() are supposed to increase the use count, iput() is
supposed to decrease the use count. When the use count is zero, the
inode is unused and may be dropped from the cache. I believe the plan
was to keep precisely one inode with zero use count in the icache at a
time.

There's _no_ locking on this code at all -- should we be refusing to
build JFFS2 if threading support is present? The core code should be
fine but fs-ecos needs some TLC.

-- 
dwmw2