[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-fsdevel
Subject:    RE: journal requirements for buffer.c (was: Roma progress report)
From:       Chris Mason <clmsys () osfmail ! isc ! rit ! edu>
Date:       1999-10-13 14:25:48
[Download RAW message or body]

We really are agreeing with each other here, I think I'm just doing a bad
job of explaining things.

bdflush could pressure the commit threads, if I had one.  Right now, bdflush
is the commit thread.  This is not hard to fix, and will be changed with the
2.3 port.

-chris

> -----Original Message-----
> From: root@idiom.com [mailto:root@idiom.com]On Behalf Of Hans Reiser
>
> Why can't bdflush pressure the commit thread(s) without waiting on them to
> complete their tasks?
>
> Why can't it pressure the FS, have the FS schedule the IO, have
> the FS tell some
> thread to wait on that I/O and complete the commit once the IO
> completes, have
> the FS return so bdflush can pressure some other owner of some
> other buffer?  If
> it flushes some extra buffers, so what, it is just a watermark....
>
> Hans
>
>

[prev in list] [next in list] [prev in thread] [next in thread]