Re: linux-2.4.10-pre5

[Linus Torvalds]

On Sun, 9 Sep 2001, Andrea Arcangeli wrote:
>
> > filesystems. The only case it doesn't like is the "rw-open of a device
> > that is rw-mounted".
>
> it also doesn't work for ro-open of a device that is rw-mounted, hdparm
> -t as said a million of times now.

It _does_ work for that case, and you just aren't reading my emails.

You only need to invalidate the device if the open was a read-write open.

It would be _stupid_ to force a writeback and device invalidate for
read-only opens, now wouldn't it?

The fact that you cannot know the difference between a read-only and a
read-write open is _entirely_ due to the fact that you leave the flush
until the last close. If you do it at every close (like I have said for
the last twohundred mails or so), you can trivially see if the open was a
read-only or not.

		Linus