libblkid: newbie question about enabling whole disk sysfs probe

libblkid: newbie question about enabling whole disk sysfs probe

[Venkatraman S]

Hi,
 I am new to util-linux and working on adding some intelligence into
mkswap to be
aware of the underlying device properties.

I want to understand how to discover a sysfs property of a disk, when
mkswap is actually
invoked with a partition id of that disk.

If a new probe is invoked with these calls set, and mkswap is invoked
on the "whole disk" /dev/sdb

       blkdid_probe pr = new_prober(fd);
       blkid_probe_enable_superblocks(pr, 1);
       blkid_probe_enable_partitions(pr, 1);
       blkid_probe_set_superblocks_flags(pr, BLKID_SUBLKS_MAGIC);
       blkid_probe_enable_topology(pr, 1);

I am able to get to the property value (which is set from the sysfs
value for the disk).

But when mkswap is invokded with /dev/sdbX, then sysfs_probe is not
able to discover the parent disk's sysfs property.

Obviously it's easy to do string manipulation and invoke the probe for
the whole disk
in the latter case, I am assuming there is a library interface that'll
do the trick.
Any help would be great.

Thanks,
Venkat.

Re: libblkid: newbie question about enabling whole disk sysfs probe

[Karel Zak]

On Mon, Oct 22, 2012 at 06:15:57PM +0530, Venkatraman S wrote:
>        blkid_probe_enable_topology(pr, 1);

This is strings based interface (NAME=value). It's probably
better to use:

   tp = blkid_probe_get_topology(pr);

   blkid_topology_get_alignment_offset(tp);

see libblkid/samples/topology.c.


> I am able to get to the property value (which is set from the sysfs
> value for the disk).
> 
> But when mkswap is invokded with /dev/sdbX, then sysfs_probe is not
> able to discover the parent disk's sysfs property.

Really? It should be able to translate the partition name to the
whole-disk name and us it.

# ./sample-topology /dev/sdd1
----- binary interface:
	alignment offset     : 0
	minimum io size      : 4096
	optimal io size      : 262144
	logical sector size  : 4096
	physical sector size : 4096
----- NAME=value interface (values: 4):
	MINIMUM_IO_SIZE = 4096
	OPTIMAL_IO_SIZE = 262144
	PHYSICAL_SECTOR_SIZE = 4096
	LOGICAL_SECTOR_SIZE = 4096



> Any help would be great.

 What exactly is your goal?

    Karel


-- 
 Karel Zak  <kzak@redhat.com>
 http://karelzak.blogspot.com

Re: libblkid: newbie question about enabling whole disk sysfs probe

[Venkatraman S]

Hi Karel,
On 22 October 2012 19:58, Karel Zak <kzak@redhat.com> wrote:
> On Mon, Oct 22, 2012 at 06:15:57PM +0530, Venkatraman S wrote:
>>        blkid_probe_enable_topology(pr, 1);
>
> This is strings based interface (NAME=value). It's probably
> better to use:
>
>    tp = blkid_probe_get_topology(pr);
>
>    blkid_topology_get_alignment_offset(tp);
>
> see libblkid/samples/topology.c.
>
>
>> I am able to get to the property value (which is set from the sysfs
>> value for the disk).
>>
>> But when mkswap is invokded with /dev/sdbX, then sysfs_probe is not
>> able to discover the parent disk's sysfs property.
>
> Really? It should be able to translate the partition name to the
> whole-disk name and us it.
>
> # ./sample-topology /dev/sdd1
> ----- binary interface:
>         alignment offset     : 0
>         minimum io size      : 4096
>         optimal io size      : 262144
>         logical sector size  : 4096
>         physical sector size : 4096
> ----- NAME=value interface (values: 4):
>         MINIMUM_IO_SIZE = 4096
>         OPTIMAL_IO_SIZE = 262144
>         PHYSICAL_SECTOR_SIZE = 4096
>         LOGICAL_SECTOR_SIZE = 4096
>
I understand. I assume this works because each of these are present in both
/sys/block/sda/ and /sys/block/sdaX/. But I am seeking a new property
(for which I have already coded the appropriate lookup hooks in topology.c)
The property (device/preferred_erase_size is present usually in eMMC
and SD cards,
 and not regular hard disks, and it only present in the root device node sysfs

I think it would be easier for you after looking at the patches that I
already have
(crude with lots of hacks and debug). I'll post them now

>
>
>> Any help would be great.
>
>  What exactly is your goal?
>

My objective is to implement / experiment some of the ideas that were bounced
around in kernel mailing list for optimizing swap on flash media [1]
[2] and make compatible
changes in both kernel and mkswap.
The changes revolve around embedding useful information in swap header that the
kernel can use to optimize it's swap algorithm.

[1] https://lkml.org/lkml/2012/3/30/410
[2] http://thread.gmane.org/gmane.linux.kernel.mmc/14864