All of lore.kernel.org
 help / color / mirror / Atom feed
From: Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org>
To: edward.robbins@domain.hid
Cc: xenomai@xenomai.org
Subject: Re: [Xenomai-help] Very high latencies under stress testing
Date: Fri, 13 Aug 2010 16:35:41 +0200	[thread overview]
Message-ID: <4C65583D.3030800@domain.hid> (raw)
In-Reply-To: <1281709057.650414331@domain.hid>

edward.robbins@domain.hid wrote:
> Hello all,
> 
> I have just set up my first xenomai system, and though I believe I
> have followed most of the guidelines, if I run latency it sits at
> around 8uS max, until I run something to stress the system (e.g.
> startx, or killing the x server) when it jumps to 700-800uS.
> 
> I wonder if this is to do with graphics? I have set "NoAccel" in the
> driver section of Xorg.conf (I am using fbdev). I also notice that
> latencies go massive (e.g. 4200uS) the first time I run glxgears.
> "glxinfo" reports "Direct rendering: no".

This is bad news. It could come either from the fact that the Xorg
driver shuts hardware interrupts off (there is nothing xenomai can do
about that), or it could be an SMI issue. What about using the driver
for the real graphic card you have ? Probably intel card? Do you observe
any big latencies if you stress the system with anything else than X?

> Power management options (ACPUI, APM): Disable ACPI (Advanced
> Configuration and Power Interface) Support --> Processor Disable CPU
> Frequency Scaling Disable CPU idle PM support Disable Power Managment
> Support

Disabling ACPI is a bad idea. As repeated many times, only
ACPI_PROCESSOR should be disabled, not ACPI.

> 
> Real-time sub-system: Enable Machine--> SMI workaround Enable
> Machine--> globally disable SMI

Do you have a message in the kernel boot logs telling you that the SMI
workaround actually worked?

> 
> I have read through this:
> http://svn.gna.org/svn/xenomai/trunk/TROUBLESHOOTING and tried to
> find anything relevant. I am booting from an on board compact flash
> via USB. I wondered if this could be the problem. Output of hdparm:

Well, this is not the latest version of the TROUBLESHOOTING guide. The
latest version is in Xenomai sources, or accessible through our website.
Though I am not sure much things have changed.


> IDE Controller settings "HDC configure as" Options: "IDE", "RAID",
> "AHCI"

Do not select IDE here, it probably uses SMIs

> 
> USB Configuration "USB Legacy" Enable/Disable "EHCI1" Enable/Disable 
> "EHCI2"            " "UHCI1"            " "UHCI2" ... "UHCI5" 
> "Per-Port Control2" Enable/Disable "USB Pre-fetch Feature"
> Enable/Disable "USB Pre-fetch Time" 2ms/4ms "USB HC Alignment"
> Enable/Disable

Do not select USB legacy, it probably uses SMI.

> 
> Chipset Configuration "CRID/SRID" Enable/Disable "Port 80h Cycles"
> LPC bus/PCI bus "DMI Link ASPM Control" Enable/Disable "Automatic
> ASMO" Auto/Manual "ASPM L0s Support" Enable/Disable "ASPM L0sL1
> Support" Enable/Disable "VT-d" Enable/Disable
> 
> ACPI Table/Feature Control "FACP - C2 Latency Value" Enable/Disable 
> "FACP - C3 Latency Value" Enable/Disable "FACP - RTC S4 Wakeup"
> Enable/Disable "APIC - IO APIC Mode" Enable/Disable "HPET - HPET
> Support" Enable/Disable

You probably want the APIC and IO APIC.

> So, could anyone suggest what I might try to improve the system?
> Sorry for the torrent of information... I thought it better to give
> as much info as I could think of than to not provide enough to make a
> diagnosis!

test whether you have issues without the X server,
if yes, check whether the SMI workaround actually works,
if no, test with another X server (the one specific to your graphic
card, presumably the intel one).

-- 
					    Gilles.


  reply	other threads:[~2010-08-13 14:35 UTC|newest]

Thread overview: 38+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2010-08-13 14:17 [Xenomai-help] Very high latencies under stress testing edward.robbins
2010-08-13 14:35 ` Gilles Chanteperdrix [this message]
2010-08-16 12:13   ` edward.robbins
2010-08-16 12:32     ` Gilles Chanteperdrix
2010-09-03 16:30       ` edward.robbins
2010-09-04  8:51         ` Philippe Gerum
2010-10-15  9:21           ` edward.robbins
2010-10-15 11:54             ` edward.robbins
2010-10-15 13:51               ` Gilles Chanteperdrix
2010-10-15 14:13                 ` edward.robbins
2010-10-15 14:25                   ` Gilles Chanteperdrix
2010-10-15 14:33                   ` Philippe Gerum
2010-10-19 13:52                 ` edward.robbins
2010-10-19 20:01                   ` Gilles Chanteperdrix
2010-10-23 11:08                     ` edward.robbins
2010-10-23 11:13                       ` Gilles Chanteperdrix
2010-10-23 11:22                         ` edward.robbins
2010-10-23 11:36                           ` Gilles Chanteperdrix
2010-10-23 13:45                             ` edward.robbins
2010-10-23 13:51                               ` Gilles Chanteperdrix
2010-10-23 13:56                                 ` edward.robbins
2010-10-23 13:58                                   ` Gilles Chanteperdrix
2010-10-23 14:32                                     ` edward.robbins
2010-10-23 14:37                                       ` Gilles Chanteperdrix
2010-10-23 14:48                                         ` edward.robbins
2010-10-23 15:16                                       ` edward.robbins
2010-10-23 14:10                                   ` Gilles Chanteperdrix
2010-08-13 14:40 ` Philippe Gerum
2010-08-13 16:15   ` Philippe Gerum
  -- strict thread matches above, loose matches on Subject: below --
2010-10-23 14:54 edward.robbins
2010-10-23 15:37 ` Gilles Chanteperdrix
2010-10-23 16:21   ` edward.robbins
2010-10-23 16:26     ` edward.robbins
2010-10-23 18:46       ` edward.robbins
2010-10-23 21:52         ` Gilles Chanteperdrix
2010-10-24 15:55           ` edward.robbins
2010-10-23 16:13 ` Gilles Chanteperdrix
2010-10-23 16:22   ` Gilles Chanteperdrix

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=4C65583D.3030800@domain.hid \
    --to=gilles.chanteperdrix@xenomai.org \
    --cc=edward.robbins@domain.hid \
    --cc=xenomai@xenomai.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.