* Subject: Multiple issues with partx
@ 2016-03-22 12:17 Serge van den Boom
2016-03-22 14:50 ` Karel Zak
0 siblings, 1 reply; 4+ messages in thread
From: Serge van den Boom @ 2016-03-22 12:17 UTC (permalink / raw)
To: util-linux
Hello,
I have encounted a few issues with 'partx'.
1. 'partx --nr 1:' does not work.
If you leave the upper range empty, partx uses 0 instead. So instead
of the list of all partitions -- which you should get according to
the manual -- you get an error:
partx: specified range <1:0> does not make sense
Changing the default to -1 in parse_range() would fix this, but see
issue #2 below.
Also note that while *upper is set to 'def' in parse_range:
*upper = *lower = def;
this value is not actually used, because a few lines below there is:
if (*end == ':' && !*(end + 1)) /* <M:> */
*upper = 0;
If the possibility to override the default value is desired, it would
imho be best to have separate default values for the lower and upper
range.
2. 'partx --nr <M:N>' does not take empty partition table entries into
account when using negative indexes.
I have created a test file with a partition table with the first and
third entry used, but the second entry open:
$ partx testfile
NR START END SECTORS SIZE NAME UUID
1 2048 10000 7953 3.9M 4fafc399-01
3 10001 20479 10479 5.1M 4fafc399-03
One might expect that 'partx --nr 1:-1' would list all partitions (i.e.
from the first to the last partition), but this is currently not true in
this case:
$ partx --nr 1:-1 testfile
NR START END SECTORS SIZE NAME UUID
1 2048 10000 7953 3.9M 4fafc399-01
In fact, 'partx --nr -1:-1' returns no output at all.
This is because negative values are offset against the number of partitions,
which is less than the number of the partition with the highest number.
One approach to solving this would be to offset against the number of
the last partition in the partition table, but this would not work as
expected when the partitions are our of order.
Another approach would be to offset negative numbers against the number
of the partition with the highest partition number.
In the case that there are holes in the partition table, this would not
necessarily return the same number of partitions as would be expected
by looking at the range. e.g. Using the same test file as above, 'partx
--nr -2:-1' would produce only one entry:
NR START END SECTORS SIZE NAME UUID
3 10001 20479 10479 5.1M 4fafc399-03
Yet another approach may be to consider the list of partitions
in the order in which they are normally shown by partx, and consider
negative numbers to be indices offset against the length of the list.
The disadvantage here is that negative numbers don't mean the
same thing as positive numbers; positive numbers refer to partition
numbers, while negative numbers refer to indices in a sorted list.
'partx --nr 1:-1' would then mean 'the partitions in the sorted list of
partitions, starting with the one with number 1, up to the last
partition in the list.
Then there are the options to add a separate option with a different
meaning for negative indices, or to remove the support for negative numbers
altogether.
3. It is not always clear whether a single partition table entry refers
to the partition table on the block device, or to a partition table
entry of the block device of which the specified block device is a part.
When the specified block device contains a partition table, that
partition table is shown. When it does not contain a partition table,
but the block device is a partition itself, its entry on its parent
block device is shown. When you don't know whether a given block device
contains is a full disk or a partition (because you're calling partx
from a script) special logic would be required. Also, nested partition
tables are theoretically possible, though unusual.
My preferred solution would be to only show the partition table on the
specified block device, unless a partition is explicitly specified.
Perhaps the behaviour could also be modified using added parameters.
As a work-around, something like 'partx --nr 1:65535' may be used if
you are interested in the partitions of the specified block device
itself. 'partx --nr 1:' would avoid the arbitrary upper bound, but does
not work because of issue 1, and 'partx --nr 1:-1' would too, but does
not work because of issue 2.
4. It is possible to specify a partition by the device name of a
partition on another block device.
The command 'partx /dev/sdb2 /dev/sda' makes no sense, but still
produces output, namely the second partition of /dev/sda.
Regards,
Serge
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: Subject: Multiple issues with partx
2016-03-22 12:17 Subject: Multiple issues with partx Serge van den Boom
@ 2016-03-22 14:50 ` Karel Zak
2016-03-22 16:14 ` Serge van den Boom
0 siblings, 1 reply; 4+ messages in thread
From: Karel Zak @ 2016-03-22 14:50 UTC (permalink / raw)
To: Serge van den Boom; +Cc: util-linux
On Tue, Mar 22, 2016 at 01:17:42PM +0100, Serge van den Boom wrote:
> 1. 'partx --nr 1:' does not work.
Fixed.
# partx --show --nr 1: /dev/sdc
NR START END SECTORS SIZE NAME UUID
1 2048 22527 20480 10M fa3974fc-40a2-402a-941a-7d26b32e126d
2 22528 43007 20480 10M 808a2e20-7109-4f2f-9727-f474f5119231
4 63488 83967 20480 10M 8fe134d1-af94-48d0-94f2-9b757da8731f
> 2. 'partx --nr <M:N>' does not take empty partition table entries into
> account when using negative indexes.
...
> One might expect that 'partx --nr 1:-1' would list all partitions (i.e.
> from the first to the last partition), but this is currently not true in
> this case:
>
> $ partx --nr 1:-1 testfile
> NR START END SECTORS SIZE NAME UUID
> 1 2048 10000 7953 3.9M 4fafc399-01
Fixed.
Note that --nr specifies partno (column NR), rather than seq.number.
Unfortunately, code mixed partno and seq.number of the partition.
# partx --show --nr 1:-1 /dev/sdc
NR START END SECTORS SIZE NAME UUID
1 2048 22527 20480 10M fa3974fc-40a2-402a-941a-7d26b32e126d
2 22528 43007 20480 10M 808a2e20-7109-4f2f-9727-f474f5119231
4 63488 83967 20480 10M 8fe134d1-af94-48d0-94f2-9b757da8731f
> In fact, 'partx --nr -1:-1' returns no output at all.
Fixed.
# partx --show --nr -1:-1 /dev/sdc
NR START END SECTORS SIZE NAME UUID
4 63488 83967 20480 10M 8fe134d1-af94-48d0-94f2-9b757da8731f
> 3. It is not always clear whether a single partition table entry refers
> to the partition table on the block device, or to a partition table
> entry of the block device of which the specified block device is a part.
>
> When the specified block device contains a partition table, that
> partition table is shown. When it does not contain a partition table,
> but the block device is a partition itself, its entry on its parent
> block device is shown. When you don't know whether a given block device
> contains is a full disk or a partition (because you're calling partx
> from a script) special logic would be required. Also, nested partition
> tables are theoretically possible, though unusual.
You have to always specify partition (-nr or devname), whole-disk is
optional. If you want to read partitioned partition (nested PT) then
you can use '-' instead of partition name:
partx --show - /dev/sdc1
Anyway, my suggestion is to use sfdisk if you want complicated things
like nested partitions.
> As a work-around, something like 'partx --nr 1:65535' may be used if
> you are interested in the partitions of the specified block device
> itself. 'partx --nr 1:' would avoid the arbitrary upper bound, but does
> not work because of issue 1, and 'partx --nr 1:-1' would too, but does
> not work because of issue 2.
--nr 1: and --nr 1:-1 works now
> 4. It is possible to specify a partition by the device name of a
> partition on another block device.
>
> The command 'partx /dev/sdb2 /dev/sda' makes no sense, but still
> produces output, namely the second partition of /dev/sda.
Fixed,
# partx --show /dev/sdx1 /dev/sdc
partx: partition and disk name do not match
I'll prepare regression test for partx....
--
Karel Zak <kzak@redhat.com>
http://karelzak.blogspot.com
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: Subject: Multiple issues with partx
2016-03-22 14:50 ` Karel Zak
@ 2016-03-22 16:14 ` Serge van den Boom
2016-03-22 16:33 ` Karel Zak
0 siblings, 1 reply; 4+ messages in thread
From: Serge van den Boom @ 2016-03-22 16:14 UTC (permalink / raw)
To: Karel Zak; +Cc: util-linux
Thanks. Fast resolution.
On Tue, 22 Mar 2016, Karel Zak wrote:
> On Tue, Mar 22, 2016 at 01:17:42PM +0100, Serge van den Boom wrote:
> >
> > 4. It is possible to specify a partition by the device name of a
> > partition on another block device.
> >
> > The command 'partx /dev/sdb2 /dev/sda' makes no sense, but still
> > produces output, namely the second partition of /dev/sda.
>
> Fixed,
>
> # partx --show /dev/sdx1 /dev/sdc
> partx: partition and disk name do not match
Strictly speaking, it should be the devices themselves which would need
to match, not their names. I could create a few symlinks, and
'partx x0 x' would work, even when x0 and x refer to unrelated devices:
ln -s /dev/sda x
ln -s /dev/sdb2 x0
Regards,
Serge
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: Subject: Multiple issues with partx
2016-03-22 16:14 ` Serge van den Boom
@ 2016-03-22 16:33 ` Karel Zak
0 siblings, 0 replies; 4+ messages in thread
From: Karel Zak @ 2016-03-22 16:33 UTC (permalink / raw)
To: Serge van den Boom; +Cc: util-linux
On Tue, Mar 22, 2016 at 05:14:36PM +0100, Serge van den Boom wrote:
> Thanks. Fast resolution.
>
> On Tue, 22 Mar 2016, Karel Zak wrote:
> > On Tue, Mar 22, 2016 at 01:17:42PM +0100, Serge van den Boom wrote:
> > >
> > > 4. It is possible to specify a partition by the device name of a
> > > partition on another block device.
> > >
> > > The command 'partx /dev/sdb2 /dev/sda' makes no sense, but still
> > > produces output, namely the second partition of /dev/sda.
> >
> > Fixed,
> >
> > # partx --show /dev/sdx1 /dev/sdc
> > partx: partition and disk name do not match
>
> Strictly speaking, it should be the devices themselves which would need
> to match, not their names. I could create a few symlinks, and
> 'partx x0 x' would work, even when x0 and x refer to unrelated devices:
>
> ln -s /dev/sda x
> ln -s /dev/sdb2 x0
This is about identification, partx does not use the device for
anything.... not sure if we have to make it more complicated.
Karel
--
Karel Zak <kzak@redhat.com>
http://karelzak.blogspot.com
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2016-03-22 16:33 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2016-03-22 12:17 Subject: Multiple issues with partx Serge van den Boom
2016-03-22 14:50 ` Karel Zak
2016-03-22 16:14 ` Serge van den Boom
2016-03-22 16:33 ` Karel Zak
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox