[PATCHv4 6/7] zswap: add flushing support

[Andrew Morton]

On Tue, 29 Jan 2013 15:40:26 -0600
Seth Jennings <sjenning at linux.vnet.ibm.com> wrote:

> This patchset adds support for flush pages out of the compressed
> pool to the swap device

I do so hate that word "flush".  Sometimes it means "writeback", other
times it means "invalidate".  And perhaps it means "copy elsewhere then
reclaim".

Please describe with great specificity what this patch actually does
with pages, and why it does it.  And where the compression factors into
this.

The code appears to take a compressed page, decompress it into
swapcache via some means.  And then, for unexplained reasons, it starts
writeback of that swapcache page.

In zswap_flush_entry() there is a comment "page is already in the swap
cache, ignore for now".  This is very interesting.  How and why does
this come about?  Does it imply that there are two copies of the same
data floating around?  If so, how come?

Preferably all the above would be understandable by reading mm/zswap.c.