Re: [PATCH v2 -mm] make swapin readahead skip over holes

[Rik van Riel]

On 01/11/2012 04:42 PM, Johannes Weiner wrote:
> On Wed, Jan 11, 2012 at 04:30:12PM -0500, Rik van Riel wrote:
> > On 01/11/2012 04:10 PM, Johannes Weiner wrote:
> > > On Wed, Jan 11, 2012 at 02:30:44PM -0500, Rik van Riel wrote:
> > > > Ever since abandoning the virtual scan of processes, for scalability
> > > > reasons, swap space has been a little more fragmented than before.
> > > > This can lead to the situation where a large memory user is killed,
> > > > swap space ends up full of "holes" and swapin readahead is totally
> > > > ineffective.
> > > >
> > > > On my home system, after killing a leaky firefox it took over an
> > > > hour to page just under 2GB of memory back in, slowing the virtual
> > > > machines down to a crawl.
> > > >
> > > > This patch makes swapin readahead simply skip over holes, instead
> > > > of stopping at them.  This allows the system to swap things back in
> > > > at rates of several MB/second, instead of a few hundred kB/second.
> > > >
> > > > The checks done in valid_swaphandles are already done in
> > > > read_swap_cache_async as well, allowing us to remove a fair amount
> > > > of code.
> > >
> > > __swap_duplicate() also checks for whether the offset is within the
> > > swap device range.  Do you think we could remove get_swap_cluster()
> > > altogether and just try reading the aligned page_cluster range?
> >
> > That is how I implemented it originally, but we need
> > to take the swap_lock so it is cleaner to implement
> > a helper function in swapfile.c :)
>
> AFAICS, it's only needed to validate the offset against si->max, but
> this too is done in __swap_duplicate().
>
> What's otherwise left is just rounding down swp_offset(entry) and
> adding 1<<  page_cluster to it, that shouldn't need the swap_lock?
>
> Am I missing something?

Good point.  I guess we could get rid of get_swap_cluster
afterall and let __swap_duplicate deal with everything.

Andrew, Mel, is that ok with you?
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/