From: Ian Campbell Subject: Re: [PATCH OSSTEST] Debian: Wait for udev devices to settle in erase-other-disks Date: Wed, 23 Sep 2015 17:11:31 +0100 Message-ID: <1443024691.10338.239.camel@citrix.com> References: <1443023316-6775-1-git-send-email-ian.campbell@citrix.com> <22018.52349.590941.29924@mariner.uk.xensource.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <22018.52349.590941.29924@mariner.uk.xensource.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Ian Jackson Cc: xen-devel@lists.xen.org List-Id: xen-devel@lists.xenproject.org On Wed, 2015-09-23 at 16:59 +0100, Ian Jackson wrote: > Ian Campbell writes ("[PATCH OSSTEST] Debian: Wait for udev devices to > settle in erase-other-disks"): > > Otherwise we, apparently, have a race between the "test -b" + dd > > against the sda1 device node being removed and end up creating a file > > called /dev/sda1 containing 32K of zeroes, which will later render > > mke2fs very confused. > > > > For some reason we seem to reliably loose the race on Jessie i386 host > > installs, but not on Wheezy or other arches on Jessie. > > > > This was tricky to diagnose because the Debian installer main-menu > > process appears to buffer the output of things it runs before logging > > them to /var/log/syslog, which means the output of "set -x" and the > > actual affect of the commands (e.g. in the kernel messages) can be > > presented in very confusing orders. Hence this patch also adds > > extensive use of logger(1) to record what it is actually doing at the > > time it does it. These logs include the pid of the command too since > > main-menu only logs its own pid, not that of the subcommand which > > gneerated the output. > > > > Lastly this patch adds an explicit test that each device it touches is > > actually still a block device afterwards. > > > > Signed-off-by: Ian Campbell > > Acked-by: Ian Jackson > > Although > > > + logger -t osstest-erase-other-disks-\$\$ "Already ran, exiting" > > +logger -t osstest-erase-other-disks-\$\$ "Running..." > > + logger -t osstest-erase-other-disks-\$\$ "Erasing \$dev" > > + logger -t osstest-erase-other-disks-\$\$ "\$dev is no > > longer a block device!" > > + logger -t osstest-erase-other-disks-\$\$ "\$dev does not exist > > or is not a block device." > > maybe > > logmsg () { logger -t osstest-erase-other-disks-\$\$ "\$@" } > > ? I wasn't confident in getting the shell-in-perl backticks correct, or rather I was too lazy to do the inevitable extra two rounds of testing it would have needed. I can do that if you want though. Ian.