[PATCH RFC] 1/2 central workspace for zlib

David Woodhouse dwmw2 at infradead.org
Mon Jun 2 11:40:07 EDT 2003


On Mon, 2003-06-02 at 16:36, Jörn Engel wrote:
> Considering that mtdblock should not be performance critical in
> production use anyway, this is a very hard problem.  What do you
> think?

mtdblock shouldn't actually be _using_ the vmalloc'd buffer for write
caching in production at all. Anyone using mtdblock in write mode for
production wants shooting.

Perhaps we could get away with allocating it only when the device is
opened for write? Even that's suboptimal since in 2.4, JFFS2 opens the
mtdblock device for write but doesn't actually _write_ to it; just gets
the appropriate MTD device and uses that directly.

-- 
dwmw2




More information about the linux-mtd mailing list