hi,

> > I must not understand what you're saying.  You say "synchronous read
> > requests can happen on the delayed-writed page and we have no way to
> > look up the corresponding buf from a page".  That suggests to me that
> > the read will return the wrong data; am I incorrect?  If not, what is
> > the problem, exactly, if synchronous read requests happen on pages
> > that have delayed writes in the I/O queues?
> 
> I'm not sure what all the hubub is all about ... don't all file data 
> requests go through UBC in any case?

did you read my replies to tls@ and bouyer@?  weren't them enough?
or am i missing the point?

my understanding is,
- page is PG_BUSY during write back.  if we keep a delayed write
  request at buffer queue forever, the page will be left PG_BUSY forever.
- getpages will wait on the PG_BUSY page.

YAMAMOTO Takashi