bug#22199: Bug in coreutils for Linux Mint?? Package coreutils_8.21-1ubuntu5.3_amd64

[Pádraig Brady]

tag 22199 notabug
close 22199
stop

On 23/12/15 19:42, Dave wrote:
> 
> On 12/21/2015 7:39 PM, Pádraig Brady wrote:
>> On 21/12/15 22:13, Dave wrote:
>>> On 12/18/2015 4:28 AM, Pádraig Brady wrote:
>>>> On 18/12/15 00:23, Dave wrote:
>>>>> df's output shows this on a standard run(df with no flags):
>>>>>
>>>>> Filesystem     1K-blocks      Used Available Use% Mounted on
>>>>> udev             1953632         4   1953628   1% /dev
>>>>> tmpfs             400436      1216    399220   1% /run
>>>>> /dev/sda1      957218452 302330152 606241340  34% /
>>>>> none                   4         0         4   0% /sys/fs/cgroup
>>>>> none                5120         0      5120   0% /run/lock
>>>>> none             2002168         8   2002160   1% /run/shm
>>>>> none              102400         4    102396   1% /run/user
>>>>> /dev/sdb1      957218452 302510240 606061252  34% /mnt/bkup/device2
>>>>>
>>>>> But when I mount an external USB disk that's on /dev/sdc1, df shows this
>>>>> output:
>>>>>
>>>>> Filesystem     1K-blocks      Used Available Use% Mounted on
>>>>> udev             1953632         4   1953628   1% /dev
>>>>> tmpfs             400436      1272    399164   1% /run
>>>>> /dev/sdc1      957218452 302334136 606237356  34% /
>>>>> none                   4         0         4   0% /sys/fs/cgroup
>>>>> none                5120         0      5120   0% /run/lock
>>>>> none             2002168         8   2002160   1% /run/shm
>>>>> none              102400         4    102396   1% /run/user
>>>>> /dev/sdb1      957218452 302498156 606073336  34% /mnt/bkup/device2
>>>>> /dev/sdc1      957218452 302564064 606007428  34% /mnt/bkup/externaldevice
>>>>>
>>>>> The filesystem of "/" has mysteriously changed (in appearance) to
>>>>> /dev/sdc1?!  All I did was mount the /dev/sdc1 device.  In fact,
>>>>> /dev/sdc1 is ONLY mounted at /mnt/bkup/externaldevice, and the mounted
>>>>> filesystem on "/" is still /dev/sda1.  I tested by writing a couple of
>>>>> files, plus mount shows the correct info:
>>>>>
>>>>> /dev/sda1 on / type ext4 (rw,errors=remount-ro)
>>>>> /dev/sdb1 on /mnt/bkup/device2 type ext4 (rw)
>>>>>
>>>>> If I unmount /mnt/bkup/externaldevice, df still shows /dev/sdc1 as the
>>>>> "/" filesystem for a small amount of time (10-20 seconds) and then it
>>>>> goes back to the correct filesystem of /dev/sda1.
>>>>>
>>>>> What is going on??  This only happened after upgrading to the package of
>>>>> coreutils mentioned in the subject, it was not doing this previously.
>>>>> Also, to note, the first two lines of the df output regarding udev and
>>>>> tmpfs did not appear until this version either.  I did see in the
>>>>> changelog that df was updated to display some additional information, so
>>>>> perhaps it has to do with that??
>>>> I see that ubuntu has backported _some_ patches from master.
>>>> Perhaps they were incomplete. Could you please try out this latest df 
>>>> binary.
>>>>
>>>>     wget http://www.pixelbeat.org/cu/df
>>>>     chmod a+x df
>>>>     ./df
>>>>
>>>> (There is a df.sig file is the same dir for gpg verification
>>>>    using the public key associated with this email address if desired)
>>>>
>>>> cheers,
>>>> Pádraig.
>>>>
>>> Pádraig, thanks for the response. The df you provided (v8.24.113-84fcb)
>>> also shows /dev/sdc1 in place of /dev/sda1 once the /dev/sdc1 device is
>>> mounted, the same as in the example in the correspondence above. Is that
>>> what you were expecting?
>>>
>>> I found a copy of the old df that was in use before the coreutils
>>> package upgrade, and it presents as version 8.21 just like the current
>>> one when running df --version. The package the original came from was
>>> coreutils_8.21-1ubuntu5.1_amd64 and the df executable is dated Jan 13,
>>> 2015. I just ran THAT df, and here's the output (with the /dev/sdc1
>>> device mounted):
>>>
>>> Filesystem 1K-blocks Used Available Use% Mounted on
>>> /dev/sda1 957218452 304082900 604488592 34% /
>>> none 4 0 4 0% /sys/fs/cgroup
>>> udev 1953632 4 1953628 1% /dev
>>> tmpfs 400436 1272 399164 1% /run
>>> none 5120 0 5120 0% /run/lock
>>> none 2002168 8 2002160 1% /run/shm
>>> none 102400 4 102396 1% /run/user
>>> /dev/sdb1 957218452 304106588 604464904 34% /mnt/bkup/device2
>>> /dev/sdc1 957218452 300703220 607868272 34% /mnt/bkup/externaldevice
>>>
>>> So, 2 things of note: 1) I was clearly incorrect in saying in my
>>> original message that the udev and tmpfs lines weren't showing in the
>>> original df listing, because there they are, they were just in a
>>> different spot (ie, not at the top) as you can see above. Apologies for
>>> the misinformation on that. 2) Well, you can see that the old df is
>>> reporting the root filesystem as /dev/sda1 correctly. If I run your df,
>>> or the newer df, both incorrectly show "/dev/sdc1" for the root
>>> filesystem after /dev/sdc1 is mounted at /mnt/bkup/externaldevice.
>>>
>>> Also, here are the relevant lines from /proc/mounts and
>>> /proc/self/mountinfo you had requested:
>>>
>>> /proc/mounts:
>>> rootfs / rootfs rw 0 0
>>> /dev/sdb1 /mnt/bkup/device2 ext4 rw,relatime,data=ordered 0 0
>>> /dev/sdc1 /mnt/bkup/externaldevice ext4 rw,relatime,data=ordered 0 0
>>>
>>> /proc/self/mountinfo:
>>> 22 1 8:1 / / rw,relatime - ext4
>>> /dev/disk/by-uuid/ced6112c-e347-4ac4-a160-0b2f03c88602
>>> rw,errors=remount-ro,data=ordered
>>> 23 22 8:17 / /mnt/bkup/device2 rw,relatime - ext4 /dev/sdb1 rw,data=ordered
>>> 33 22 8:33 / /mnt/bkup/externaldevice rw,relatime - ext4 /dev/sdc1
>>> rw,data=ordered
>>>
>>> Also, it should be noted, upon looking further into the uuid for "/"
>>> referenced in /proc/self/mountinfo, I found that that reference is
>>> softlinked to "/dev/sdc1" and it stays that way even after the external
>>> device is unmounted. Hmm. So /proc/self/mountinfo seems wrong about "/"
>>> in this case, since it's not referencing /dev/sda in any way. And it's
>>> odd that the older df works, even despite that issue. Maybe I'm not
>>> clear on what all /proc/self/mountinfo is supposed to convey? And maybe
>>> this bug is originating somewhere else outside of df then?
>> Neither /proc/self/mounts or /proc/self/mountinfo mention /dev/sda1 ?
>> That's weird. Note the older ubuntu df should be referencing /etc/mtab
>> which these days is usually a link to /proc/self/mounts.
>> The newer df will directly reference /proc/self/mountinfo.
>>
>> So it would be good to confirm that /etc/mtab is a symlink to 
>> /proc/self/mounts.
>> Also it would be good to attach that file and /proc/self/mountinfo
>> to avoid any ambiguity.
>>
>> thanks,
>> Pádraig
>>
> Can't get those files attached at the moment, but I can confirm that 
> /etc/mtab is NOT a symlink to anything and is in fact its own file. And 
> this non-symlink /etc/mtab does reference /dev/sda1.  FWIW.

OK, well that explains why the older df works, as it only
references the static /etc/mtab with no reference to
either /proc/self/mounts or /proc/self/mountinfo.
On your system these kernel generated files are incorrect for the / entry
after mounting /dev/sdc1, and for about 20 seconds after unmounting that device.
The newer df is only propagating that incorrect information,
and so this seems to be a kernel or udev issue on your system.

thanks,
Pádraig.