Re: [PATCH] writeback: Avoid exhausting allocation reserves under memory pressure

[Michal Hocko <mhocko@kernel.org>]

On Thu 05-05-16 11:07:50, Jan Kara wrote:
> On Thu 05-05-16 10:24:33, Michal Hocko wrote:
> > > +/*
> > > + * Check whether the request to writeback some pages can be merged with some
> > > + * other request which is already pending. If yes, merge it and return true.
> > > + * If no, return false.
> > > + */
> > > +static bool wb_merge_request(struct bdi_writeback *wb, long nr_pages,
> > > +			     struct super_block *sb, bool range_cyclic,
> > > +			     enum wb_reason reason)
> > > +{
> > > +	struct wb_writeback_work *work;
> > > +	bool merged = false;
> > > +
> > > +	spin_lock_bh(&wb->work_lock);
> > > +	list_for_each_entry(work, &wb->work_list, list) {
> > 
> > Is the lenght of the list bounded somehow? In other words is it possible
> > that the spinlock would be held for too long to traverse the whole list?
> 
> I was thinking about this as well. With the merging enabled, the number of
> entries queued from wb_start_writeback() is essentially limited by the
> number of writeback reasons and there's only a couple of those. What is
> more questionable is the number of entries queued from
> __writeback_inodes_sb_nr(). Generally there should be a couple at maximum
> either but it is hard to give any guarantee since e.g. filesystems use this
> function to reduce amount of delay-allocated data when they are running out
> of space. Hum, maybe we could limit the merging to scan only the last say
> 16 entries. That should give good results in most cases... Thoughts?

Yes, I guess this sounds reasonable. The merging doesn't have to be
perfect. We primarily want to get rid of (potentially) thousands of
duplicates. And there is always a GFP_NOWAIT fallback IIUC.
-- 
Michal Hocko
SUSE Labs

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>