Re: reiserfsprogs 3.6.5-pre2 release.

[Oleg Drokin <green@namesys.com>]

Hello!

On Wed, Feb 26, 2003 at 12:21:44PM -0700, Andreas Dilger wrote:

> > tests shown that almost all the time was spent on reading bitmap blocks.
> > And the same reading is done at mount time actually.
> It would be interesting to see if reiserfsck immediately followed by mount
> (which is the normal boot behaviour) takes any longer than just the mount.

Two subsequent fsck -a takes twice the time. Almost (+- 2 secs of 48)

> Oleg posted that the block device cache is flushed upon last close, but I
> don't think that is totally true.  It _was_ true for a short while, but

It is. You can trivially verify that.

> it was rather annoying and I think Al Viro fixed that so that the block
> device cache was "lazily" dropped, so you would keep it if you re-opened
> the device shortly after closing it.

No, I do not think so. Invalidate inode pages (or whatever) is called on
last device close. Anyway think of two fscks on two filesystems, then mounting.

Though last time I played with stock SuSE kernels (not so long ago), they had
this behavior turned off and cache survived device close.

> I suppose the other way to test it would be to run 2 "reiserfsck -a"
> in a row.

Yes, Philippe did this for us already.

Bye,
    Oleg