From owner-freebsd-hackers Tue Aug 8 9: 4:45 2000 Delivered-To: freebsd-hackers@freebsd.org Received: from guardian.hermes.si (guardian.hermes.si [193.77.5.150]) by hub.freebsd.org (Postfix) with ESMTP id DD5F237B7E8 for ; Tue, 8 Aug 2000 09:04:39 -0700 (PDT) (envelope-from mitja@hermes.si) Received: from hermes.si (primus.hermes.si [193.77.5.98]) by guardian.hermes.si (8.9.3/8.9.3) with ESMTP id SAA06442 for ; Tue, 8 Aug 2000 18:04:31 +0200 (METDST) Received: (from uucp@localhost) by hermes.si (8.9.3/8.9.3) id SAA20206 for ; Tue, 8 Aug 2000 18:04:27 +0200 Received: from akira.hermes.si(10.17.1.122) by primus.hermes.si via smap (V2.1) id xma019113; Tue, 8 Aug 00 18:03:10 +0200 Received: from lamu.hermes.si (lamu.hermes.si [10.17.1.230]) by akira.hermes.si with ESMTP (8.8.6 (PHNE_17135)/8.7.1) id SAA11004 for ; Tue, 8 Aug 2000 18:06:21 +0200 (METDST) Message-Id: <200008081606.SAA11004@akira.hermes.si> Date: Tue, 08 Aug 2000 18:03:08 CEST From: Mitja Horvat To: freebsd-hackers@freebsd.org Subject: 4.1-RELEASE problem writing to async mounted filesystem Reply-To: mitja.horvat@hermes.si X-Mailer: Spruce 0.6.5 for X11 w/smtpio 0.7.9 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 8bit Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG Hi, I recently upgraded to FreeBSD 4.1-RELEASE(CD image). I noticed that while writing to an asynchronously mounted filesystem(mount -o async / ...) all other IO operations to the FS are almost blocked. I have only one asynchronously mounted partition(/) and the following shell script makes the machine unresponsive so I cannot even telnet to it: while true; do rm -f foo; dd if=/dev/zero of=foo bs=1024k count=256; done 4.0-RELEASE does a much better job at handling similar situations... I'm using softupdates, but even without them the problem persists. This does not happen if the fs is mounted synchronously, but I also get MUCH slower IO throughput(800kb/s compared to 12MB/s with async). I know that mounting filesystems asynchronously is dangerous, but this is my workstation and is being backed up on a daily basis, so a filesystem crash is not something I'm worried about... I suspect this is something related to the buffer code and currently have no much time to inspect that. Can somebody give me a hint how to solve the problem? Thank you, Mitja To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message