GNU bug report logs - #21272
partprobe does not wait for refresh of partition info

Previous Next

To reply to this bug, email your comments to 21272 AT debbugs.gnu.org.

Toggle the display of automated, internal messages from the tracker.

Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):

From: Bjoern Eberth <bjoern.eberth <at> gmail.com>
To: bug-parted <at> gnu.org
Subject: partprobe does not wait for refresh of partition info
Date: Sun, 16 Aug 2015 01:22:15 +0200

[Message part 1 (text/plain, inline)]

Hi all,

As the web bugtracker seems to be broken I write this mail.

I'm using parted 3.2. All this was tested running a Debian jessie release
within a virtual machine on Windows 8.1 using VMware Workstation 11.
/dev/sdb is an USB cardreader physically connected to the Windows 8.1
machine and looped to the virtual machine.

I was executing a script that does 'partprobe /dev/sdb' and immediatly
after that 'dd of=/dev/sdb1 if=<something>'. The result is, that I get a
regular file /dev/sdb1.

After some more investigation I found, that after executing partprobe the
partition files got removed and readded to the system.
So running the script:
root <at> debian:~# cat /tmp/test
partprobe /dev/sdb
while true; do
        ls -lad /dev/sdb1
done

results in:

root <at> debian:~# /tmp/test
brw------- 1 root root 8, 17 Aug 15 16:39 /dev/sdb1
ls: cannot access /dev/sdb1: No such file or directory
ls: cannot access /dev/sdb1: No such file or directory
ls: cannot access /dev/sdb1: No such file or directory
ls: cannot access /dev/sdb1: No such file or directory
ls: cannot access /dev/sdb1: No such file or directory
brw------- 1 root root 8, 17 Aug 15 16:39 /dev/sdb1
brw------- 1 root root 8, 17 Aug 15 16:39 /dev/sdb1
brw------- 1 root root 8, 17 Aug 15 16:39 /dev/sdb1
[...]
brw-rw---- 1 root disk 8, 17 Aug 15 16:39 /dev/sdb1
brw-rw---- 1 root disk 8, 17 Aug 15 16:39 /dev/sdb1
brw-rw---- 1 root disk 8, 17 Aug 15 16:39 /dev/sdb1^C

All the above is happing in way less then a second after executing
partprobe.
Thats a really odd behaviour.

Then I downgraded parted to version 2.2 and ran the test script again:
root <at> debian:~# /tmp/test
brw------- 1 root root 8, 17 Aug 15 14:38 /dev/sdb1
brw------- 1 root root 8, 17 Aug 15 14:38 /dev/sdb1
brw------- 1 root root 8, 17 Aug 15 14:38 /dev/sdb1
brw------- 1 root root 8, 17 Aug 15 14:38 /dev/sdb1
brw------- 1 root root 8, 17 Aug 15 14:38 /dev/sdb1
brw------- 1 root root 8, 17 Aug 15 14:38 /dev/sdb1
brw------- 1 root root 8, 17 Aug 15 14:38 /dev/sdb1
[...]
brw-rw---- 1 root disk 8, 17 Aug 15 16:39 /dev/sdb1
brw-rw---- 1 root disk 8, 17 Aug 15 16:39 /dev/sdb1
brw-rw---- 1 root disk 8, 17 Aug 15 16:39 /dev/sdb1^C

This looks rather fine to me.

Is this a bug? Or is partprobe 3.2 just running much faster than partprobe
2.2 giving the kernel no time to reread the partition table? Or something
else?

Kind regards,
Bjoern

[Message part 2 (text/html, inline)]

Message #8 received at submit <at> debbugs.gnu.org (full text, mbox):

From: "Brian C. Lane" <bcl <at> redhat.com>
To: bug-parted <at> gnu.org
Subject: Re: bug#21272: partprobe does not wait for refresh of partition info
Date: Mon, 17 Aug 2015 08:58:19 -0700

On Sun, Aug 16, 2015 at 01:22:15AM +0200, Bjoern Eberth wrote:
> Hi all,
> 
> As the web bugtracker seems to be broken I write this mail.
> 
> I'm using parted 3.2. All this was tested running a Debian jessie release
> within a virtual machine on Windows 8.1 using VMware Workstation 11.
> /dev/sdb is an USB cardreader physically connected to the Windows 8.1
> machine and looped to the virtual machine.
> 
> I was executing a script that does 'partprobe /dev/sdb' and immediatly
> after that 'dd of=/dev/sdb1 if=<something>'. The result is, that I get a
> regular file /dev/sdb1.
> 
> After some more investigation I found, that after executing partprobe the
> partition files got removed and readded to the system.
> So running the script:
> root <at> debian:~# cat /tmp/test
> partprobe /dev/sdb
> while true; do
>         ls -lad /dev/sdb1
> done

udev is responsible for making sure the device nodes are correct. You
can't depend on them being present without checking for them first. A
udevadm settle after partprobe may also help.

-- 
Brian C. Lane | Anaconda Team | IRC: bcl #anaconda | Port Orchard, WA (PST8PDT)

Message #11 received at 21272 <at> debbugs.gnu.org (full text, mbox):

From: Phil Susi <psusi <at> ubuntu.com>
To: "Brian C. Lane" <bcl <at> redhat.com>, 21272 <at> debbugs.gnu.org
Subject: Re: bug#21272: partprobe does not wait for refresh of partition info
Date: Fri, 18 Sep 2015 15:59:09 -0400

On 8/17/2015 11:58 AM, Brian C. Lane wrote:
> udev is responsible for making sure the device nodes are correct. You
> can't depend on them being present without checking for them first. A
> udevadm settle after partprobe may also help.

IIRC, libparted contacts udev to wait for the dev node to be created so
this shouldn't happen.  At least if it was configured with udev.

Message #14 received at 21272 <at> debbugs.gnu.org (full text, mbox):

From: "Brian C. Lane" <bcl <at> redhat.com>
To: Phil Susi <psusi <at> ubuntu.com>
Cc: 21272 <at> debbugs.gnu.org
Subject: Re: bug#21272: partprobe does not wait for refresh of partition info
Date: Wed, 23 Sep 2015 16:40:47 -0700

On Fri, Sep 18, 2015 at 03:59:09PM -0400, Phil Susi wrote:
> On 8/17/2015 11:58 AM, Brian C. Lane wrote:
> > udev is responsible for making sure the device nodes are correct. You
> > can't depend on them being present without checking for them first. A
> > udevadm settle after partprobe may also help.
> 
> IIRC, libparted contacts udev to wait for the dev node to be created so
> this shouldn't happen.  At least if it was configured with udev.
> 

That's only done for device-mapper devices. udev (especially
systemd-udev) can trigger async. actions so you have to watch out for
how parted is being called. See my patch for opening things read-only
initially for more details.

-- 
Brian C. Lane | Anaconda Team | IRC: bcl #anaconda | Port Orchard, WA (PST8PDT)

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.