From owner-freebsd-current@FreeBSD.ORG Thu Sep 8 18:58:08 2005 Return-Path: X-Original-To: freebsd-current@freebsd.org Delivered-To: freebsd-current@freebsd.org Received: by hub.freebsd.org (Postfix, from userid 618) id D4DF416A420; Thu, 8 Sep 2005 18:58:08 +0000 (GMT) In-Reply-To: <1126200330.11226.1.camel@leguin> from Eric Anholt at "Sep 8, 2005 10:25:30 am" To: eta@lclark.edu (Eric Anholt) Date: Thu, 8 Sep 2005 18:58:08 +0000 (GMT) X-Mailer: ELM [version 2.4ME+ PL54 (25)] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Message-Id: <20050908185808.D4DF416A420@hub.freebsd.org> From: wpaul@FreeBSD.ORG (Bill Paul) Cc: truckman@freebsd.org, freebsd-current@freebsd.org, kabaev@gmail.com Subject: Re: Odd performance problem (hitching) X-BeenThere: freebsd-current@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Discussions about the use of FreeBSD-current List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Thu, 08 Sep 2005 18:58:09 -0000 > > > > > >>That could be an important clue. Maybe one of the X apps that you are > > >>running, like your mail reader, browser, or system status monitor. Try > > >>running X with one of the lightweight window managers and just an xterm > > >>or two. > > >> > > >>The 10 second interval doesn't make it sound like the problem is any of > > >>the built in kernel tasks. It's more consistent with something that > > >>runs every 10 seconds in userland that monopolizes some kernel resource > > >>whenever it runs. > > > > > > > > > Possibly, but I would expect to see some evidence in top of this. > > > > > > Also, during the 'lag' X drops or doubles up keypresses - it would suprise me > > > to find that a userland app could make X do that very easily. > > > > > > I will try your suggestion though. > > > > I had a similar problem with XFCE4's battery/temp monitor applet. I > > think my freeze was about every 6-10seconds. > > I've also had a problem with gnome's battery monitor producing long > "hitches" every several seconds. *sigh* The problem is the ACPI thermal monitor thread. Once every 10 seconds, it queries the thermal state of the system, and reading this state seems to chew up a lot of cycles with interrupts blocked. You can shut it up by editing loader.conf and adding: debug.acpi.disable="thermal" But the correct thing to do is fix the code that access ACPI resources/registers/whatever so that it doesn't block for so long. I'm really very annoyed that nobody has bothered to address this yet. -Bill -- ============================================================================= -Bill Paul (510) 749-2329 | Senior Engineer, Master of Unix-Fu wpaul@windriver.com | Wind River Systems ============================================================================= you're just BEGGING to face the moose =============================================================================