Re: RFC on io-stalls patch

[Chris Mason]

On Fri, 2003-07-11 at 20:20, Nick Piggin wrote:

> >Seems that way.  With the 2.4.21 code, a read might easily get a
> >request, but then spend forever waiting for a huge queue of merged
> >writes to get to disk.
> >
> 
> But it is the job of the io scheduler to prevent this, isn't it?
> 

Yes, but the 2.4.21 code doesn't do it.

> >
> >I believe the new way provides better overall read performance in the
> >presence of lots of writes.
> >
> >
> 
> I don't know how that can be, considering writers will consume
> basically limitless requests. What am I missing?

There is a limit on the total amount of IO in flight (4MB by default,
reads/writes combined).  We can make this a harder limit by disallowing
merges on any requests present at the time of an unplug.  Perhaps I'm
not reading your question correctly?

-chris