pdflush -> noflushd related question

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi.

I want to fix the userspace noflushd daemon to run on 2.5 kernels, but it seems to get *very* tricky, because bdflush, kupdated have been replaced by pdflush.

I've grepped through the lkml-archives and I found a post from you, Andrew, that sayed:

[SNIP]
Subject: [patch] replace kupdate and bdflush with pdflush

- - use a timer to kick off a pdflush thread every five seconds to run the kupdate code.

- - wakeup_bdflush() kicks off a pdflush thread to run the current bdflush function.

There's some loss of functionality here - the ability to tune the writeback periods. The numbers are hardwired at present. But the intent is that buffer-based writeback disappears altogether. New mechanisms for tuning the writeback will need to be introduced.
[SNIP]

Yea, but it seems, that I need exactly this lost functionality. :) noflushd, on 2.4 kernels, prevents the disks to spin up by spinning them down an stopping the kupdated daemon through putting the update-interval to 0 via syscall bdflush().
That's the code to do so:
bdflush(2 + (1 << 3) + 1, interval);

So, how to set the interval, or better sayed, how to _stop_ buffer flushing in 2.5?
Has this lost functionality already been re-implemented? Are there new syscalls?

Thanks for your help.

- --
Regards Michael Büsch
http://www.8ung.at/tuxsoft
18:29:24 up 4:01, 2 users, load average: 1.11, 1.09, 1.05

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)

iD8DBQE+2NtXoxoigfggmSgRArisAJ9KTnvSW51e4OnLvkvjI/D/bUF19ACeK9Yw xOVOelmYmHd9eC7Unuys8pM=
=t+79
-----END PGP SIGNATURE-----

/proc/sys/vm has the appropriate tunables. They are documented in Documentation/filesystems/proc.txt.

You can turn these guys off by setting the sysctls to 1000000000
I guess. Problem is, I don't think there's a way of starting them
again until the ten million seconds expires. hmm.

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Thanks Andrew, that's a good point to goon hacking. :) I'll look at it.

- --
Regards Michael Büsch
http://www.8ung.at/tuxsoft
20:21:50 up 5:54, 2 users, load average: 2.01, 2.02, 2.00

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)

iD8DBQE+2PN0oxoigfggmSgRArlBAJ9zZBsnnAjTA8s6cTK7xPeFDM+v+wCffyE/ fLHzyat7wxgEu1902BYQtUg=
=eq9e
-----END PGP SIGNATURE-----

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

I've thought a little bit more about it.
Why do you think, there is no way of waking up?
Is it, because, when I set it to 1000000000 and the back to, let's say, 500, the pdflush threads don't wake up to recognize this change? Is this the cause?

What about signaling all pdflush threads with, for example, for(ALL_PDFLUSHS)

	kill(pid, SIGSTOP);

Don't they wake up then and recognize the reducing of the timeout? The old noflushd did so to wake up updated.

thanks.

- --
Regards Michael Büsch
http://www.8ung.at/tuxsoft
20:32:32 up 6:04, 2 users, load average: 2.05, 2.03, 2.00

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)

iD8DBQE+2PZHoxoigfggmSgRAjx2AJ9orNNk8SLQlhBg+lJ1ZsyGgOb/9ACgjBjW bm0WPZwvb8Rd0bGti2XFBOQ=
=Et8g
-----END PGP SIGNATURE-----

Yes.

No, that's OK - I'll fix it in the kernel. It's pretty simple.