From: Andi Kleen <ak@suse.de>
To: "Jeff V. Merkey" <jmerkey@vger.timpanogas.org>
Cc: Andi Kleen <ak@suse.de>, linux-kernel@vger.kernel.org
Subject: Re: 2.2.18Pre Lan Performance Rocks!
Date: Mon, 30 Oct 2000 09:16:15 +0100 [thread overview]
Message-ID: <20001030091615.A610@gruyere.muc.suse.de> (raw)
In-Reply-To: <39FCB09E.93B657EC@timpanogas.org> <E13q2R7-0006S7-00@the-village.bc.nu> <20001029183531.A7155@vger.timpanogas.org> <20001030074700.A31783@gruyere.muc.suse.de> <20001029235821.A19076@vger.timpanogas.org> <20001030080858.A32204@gruyere.muc.suse.de> <20001030001646.A19136@vger.timpanogas.org> <20001030083827.B32389@gruyere.muc.suse.de> <20001030010434.A19615@vger.timpanogas.org>
In-Reply-To: <20001030010434.A19615@vger.timpanogas.org>; from jmerkey@vger.timpanogas.org on Mon, Oct 30, 2000 at 01:04:34AM -0700
On Mon, Oct 30, 2000 at 01:04:34AM -0700, Jeff V. Merkey wrote:
> > >
> > > When you say it reloads it's VM, you mean it reloads the CR3 register?
> >
> > Yes.
> >
> > No. In 2.4 you could probably use the on demand lazy vm mechanism ingo described for
> > the nfsd processes. In 2.2 it is a bit more tricky, if I remember right lazy mm needed
> > quite a few changes.
> >
> > But before doing too many changes i would first verify if that is really the problem.
>
> We will never beat NetWare on scaling if this is the case, even in 2.4.
> Andre and my first job will be to create an arch port with MANOS that
> disables this and restructures the VM.
I just guess the end result will be as crash prone as Netware when you install any third
party software ;)
lazy mm is probably a better path, as long as you stay in kernel threads and a single user mm
it'll never switch VMs.
>
> > PTEs are read for aging on memory pressure in vmscan.
> >
> > segment register reload happen on interrupt entry, to load the kernel cs/ds (are they that
> > costly?). If it was really costly you could probably check in interrupt entry if you're
> > already running in kernel space and skip it.
>
> They are. segment register reloads will trigger the following:
>
> IDT table atomic fetch to verify (LOCK#) (if triggered by task gate from INTR)
> GDT table atomic fetch to verify (LOCK#)
> LDT table atomic fetch to verify (LOCK#) (if present)
> PDE table atomic fetch to verify (LOCK#)
>
> The process has to verify that the loaded segment descriptor is valid, and
> it will fetch from all these tables to do it, with up to 4 (LOCK#)
> assertions occurring invisibly in the hardware underneath (which will
> generate 4 non-cacheable memory references, in addition to wrecking
> havoc on the affected L1/L2 cache lines). Oink. It only does this
> when you load one, not when you save one, like pushing it on the
> stack. Since you look at MANOS code, you'll note that in
> CONTEXT.386, I do and add esp, 3 * 4 instead of poppping the segment
> registers off the stack if they are in the kernel address space.
>
> Linux should do the same, if possible as an optimization.
Interesting. You could do easily the same in Linux by changing (in 2.2) the SAVE_ALL macro
in arch/i386/kernel/irq.h and doing the same in ret_from_intr's RESTORE_ALL macro (after
changing it to a RESTORE_ALL_INT or you'll break system calls)
Actually I don't even know why irq.h's SAVE_ALL even loads __KERNEL_CS, it should be already
set by the interrupt gate. __KERNEL_DS probably needs to be still loaded, but only when
you came from user space.
-Andi
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/
next prev parent reply other threads:[~2000-10-30 8:16 UTC|newest]
Thread overview: 152+ messages / expand[flat|nested] mbox.gz Atom feed top
2000-10-29 23:19 2.2.18Pre Lan Performance Rocks! Jeff V. Merkey
[not found] ` <E13q2R7-0006S7-00@the-village.bc.nu>
2000-10-30 1:35 ` Jeff V. Merkey
2000-10-30 6:47 ` Andi Kleen
2000-10-30 6:58 ` Jeff V. Merkey
2000-10-30 7:08 ` Andi Kleen
2000-10-30 7:16 ` Jeff V. Merkey
2000-10-30 7:38 ` Andi Kleen
2000-10-30 8:04 ` Jeff V. Merkey
2000-10-30 8:16 ` Andi Kleen [this message]
2000-10-30 12:47 ` Alan Cox
2000-10-30 12:50 ` Andi Kleen
2000-10-30 8:26 ` Ingo Molnar
2000-10-30 7:20 ` Jeff V. Merkey
2000-10-30 8:39 ` Ingo Molnar
2000-10-30 8:08 ` Jeff V. Merkey
2000-10-30 9:52 ` Ingo Molnar
2000-10-30 8:55 ` Jeff V. Merkey
2000-10-30 10:13 ` Ingo Molnar
2000-10-30 9:11 ` Jeff V. Merkey
2000-10-30 10:41 ` Ingo Molnar
2000-10-30 9:33 ` Jeff V. Merkey
2000-10-30 10:56 ` Ingo Molnar
2000-10-30 9:45 ` Jeff V. Merkey
2000-10-30 11:04 ` Ingo Molnar
2000-10-30 9:56 ` Jeff V. Merkey
2000-10-30 11:13 ` Ingo Molnar
2000-10-30 10:08 ` Jeff V. Merkey
2000-10-30 17:41 ` Andrea Arcangeli
2000-10-30 17:58 ` Chris Evans
2000-10-30 18:01 ` Jeff V. Merkey
2000-10-30 18:21 ` Andrea Arcangeli
2000-10-30 17:59 ` Jeff V. Merkey
2000-10-31 8:08 ` Ingo Molnar
2000-10-31 20:04 ` Jeff V. Merkey
2000-10-30 19:11 ` Dan Hollis
2000-10-31 18:59 ` Pavel Machek
2000-10-30 10:27 ` Ingo Molnar
2000-10-30 9:20 ` Jeff V. Merkey
2000-10-30 10:44 ` Ingo Molnar
2000-10-30 9:38 ` Jeff V. Merkey
2000-10-30 11:01 ` Ingo Molnar
2000-10-30 9:54 ` Jeff V. Merkey
2000-10-30 11:12 ` Ingo Molnar
2000-10-30 10:06 ` Jeff V. Merkey
2000-10-30 10:56 ` john slee
2000-10-30 18:04 ` Jeff V. Merkey
2000-10-30 11:31 ` Ingo Molnar
2000-10-30 12:57 ` Alan Cox
2000-10-30 17:55 ` Jeff V. Merkey
2000-10-30 18:34 ` Alan Cox
2000-10-30 21:17 ` Jeff V. Merkey
2000-10-31 9:25 ` Erik Andersen
2000-10-31 18:50 ` Pavel Machek
2000-10-31 20:06 ` Jeff V. Merkey
2000-10-31 20:13 ` Jeff V. Merkey
2000-10-31 21:31 ` Ingo Molnar
2000-10-31 21:56 ` Ingo Molnar
2000-10-31 21:57 ` Jeff V. Merkey
2000-11-01 0:27 ` Ingo Molnar
2000-10-31 23:18 ` Jeff V. Merkey
2000-11-01 0:47 ` Ingo Molnar
2000-11-01 0:56 ` Davide Libenzi
[not found] ` <20001102031546.B10806@cerebro.laendle>
[not found] ` <20001101212835.A2425@vger.timpanogas.org>
[not found] ` <20001102043332.A27126@fuji.laendle>
[not found] ` <3A0195DA.DDEBAC51@timpanogas.org>
[not found] ` <20001102194323.D2790@cerebro.laendle>
[not found] ` <3A01CBB5.48C3094A@timpanogas.org>
[not found] ` <20001102214903.F2790@cerebro.laendle>
[not found] ` <3A01E71A.778BD898@timpanogas.org>
[not found] ` <20001102232210.H2790@cerebro.laendle>
[not found] ` <3A01ECD2.76DE10FF@timpanogas.org>
2000-11-02 22:46 ` Jeff V. Merkey
2000-11-03 0:12 ` Davide Libenzi
2000-11-02 23:00 ` Jeff V. Merkey
2000-10-31 21:34 ` Ingo Molnar
2000-10-31 21:52 ` Jeff V. Merkey
2000-10-31 22:05 ` Andi Kleen
2000-10-31 22:23 ` Jeff V. Merkey
2000-10-31 22:45 ` Jeff V. Merkey
2000-10-31 22:44 ` David Lang
2000-10-31 22:57 ` Jeff V. Merkey
2000-10-31 22:52 ` David Lang
2000-10-31 23:02 ` Alan Cox
2000-10-31 23:03 ` Jeff V. Merkey
2000-11-01 0:08 ` Ingo Molnar
2000-10-31 22:59 ` Jeff V. Merkey
2000-11-01 2:30 ` Horst von Brand
2000-10-31 23:05 ` Richard B. Johnson
2000-10-31 23:14 ` Jeff V. Merkey
2000-11-01 0:32 ` Ingo Molnar
2000-10-31 23:23 ` Jeff V. Merkey
2000-11-01 0:55 ` Ingo Molnar
2000-10-31 23:54 ` Ingo Molnar
2000-10-31 22:47 ` Jeff V. Merkey
2000-10-31 22:56 ` Larry McVoy
2000-10-31 22:55 ` Jeff V. Merkey
2000-11-01 0:10 ` Ingo Molnar
2000-11-01 5:38 ` Daniel Phillips
2000-11-03 6:42 ` Daniel Phillips
2000-10-31 22:59 ` Michael H. Warfield
2000-10-31 23:12 ` Ingo Molnar
2000-10-31 22:28 ` Jeff V. Merkey
2000-11-01 5:01 ` Peter Samuelson
2000-11-01 5:09 ` Larry McVoy
2000-11-01 5:20 ` Peter Samuelson
2000-10-30 10:50 ` Ingo Molnar
2000-10-30 9:40 ` Jeff V. Merkey
2000-10-30 23:26 ` David Woodhouse
2000-10-30 23:49 ` Jeff V. Merkey
2000-10-31 23:34 ` Roger Larsson
2000-10-31 15:18 ` Reto Baettig
2000-10-31 20:26 ` Alan Cox
2000-10-31 15:30 ` Reto Baettig
2000-10-31 20:37 ` Alan Cox
2000-10-31 20:48 ` Jesse Pollard
2000-10-31 20:58 ` Alan Cox
2000-11-01 1:33 ` Horst von Brand
2000-11-01 3:42 ` Jesse Pollard
2000-11-01 13:26 ` Horst von Brand
2000-10-31 21:43 ` Jeff V. Merkey
2000-10-31 21:50 ` Jeff V. Merkey
2000-10-31 20:36 ` Rik van Riel
2000-10-31 15:47 ` Reto Baettig
2000-10-31 21:05 ` Rik van Riel
2000-10-31 21:33 ` Jeff V. Merkey
2000-10-31 21:48 ` Rik van Riel
2000-10-31 16:54 ` Reto Baettig
2000-10-31 21:58 ` Rik van Riel
2000-10-31 21:53 ` Jeff V. Merkey
[not found] <200010300823.BAA19834@vger.timpanogas.org>
2000-10-30 8:36 ` Jeff V. Merkey
2000-10-31 21:44 Jeff V. Merkey
2000-10-31 21:36 ` Paul Menage
2000-10-31 21:44 ` David S. Miller
2000-10-31 23:21 ` Matti Aarnio
2000-10-31 23:39 ` David Weinehall
2000-10-31 22:05 ` Larry McVoy
2000-10-31 22:15 ` Jeff V. Merkey
2000-10-31 22:27 ` Larry McVoy
2000-10-31 22:38 ` Jeff V. Merkey
2000-10-31 22:48 ` Rik van Riel
2000-10-31 22:50 ` Jeff V. Merkey
2000-10-31 22:49 ` Larry McVoy
2000-10-31 22:53 ` Jeff V. Merkey
2000-10-31 23:15 ` Nathan Paul Simons
2000-10-31 23:20 ` Jeff V. Merkey
2000-11-01 0:00 ` Michael H. Warfield
2000-11-01 0:07 ` Alan Cox
2000-11-01 0:13 ` Michael H. Warfield
2000-11-01 1:30 ` Andrea Arcangeli
[not found] ` <20001031183809.C9733@.timpanogas.org>
2000-11-01 15:41 ` Andrea Arcangeli
2000-11-01 17:25 ` Jeff V. Merkey
2000-11-01 17:35 ` Jeff V. Merkey
2000-11-01 18:07 ` Andrea Arcangeli
2000-11-01 18:34 ` Jeff V. Merkey
2000-11-01 21:13 ` Daniel Phillips
2000-11-01 21:32 ` Jeff V. Merkey
2000-11-01 11:13 ` David Woodhouse
2000-11-01 15:00 ` Andrea Arcangeli
2000-11-01 9:51 ` Anton Altaparmakov
2000-11-01 17:28 ` Jeff V. Merkey
2000-11-02 21:58 ` Anton Altaparmakov
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20001030091615.A610@gruyere.muc.suse.de \
--to=ak@suse.de \
--cc=jmerkey@vger.timpanogas.org \
--cc=linux-kernel@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).