I hate to bitch but bitch I must

[Vincent Hoffman]

Polytropon wrote:
> On Sat, 17 Oct 2009 23:17:29 +0100, Vincent Hoffman <vince at unsane.co.uk> wrote:
>   
>> yes. this makes a ufs label which you can access via /dev/ufs
>> for example (my home system)
>> jhary at ostracod
>> (23:08:34 <~>) 0 $ ls /dev/ufs
>> SCRATCH         SSDROOT         SSDUSR          SSDVAR
>> [...]
>> /dev/ufs/SCRATCH on /scratch (ufs, local, noatime, gjournal)
>>     
>            ^^^^^^^    ^^^^^^^^
>
> Wow! Last time I saw this was on EAW's WEGA (a UNIX System III
> compatible UNIX developed in the GDR for the P8000 workstation).
> There even was /etc/mount and /etc/fsck. :-)
>
>
>   
Heh nothing so arcane here, I just use it as a scratch (disposable
content) disk.
>> /dev/ufs/SSDVAR /var    ufs     rw,noatime      2       2
>> /dev/label/SWAP none    swap    sw      0       0
>>     
>
> These two lines illustrate the different use of the results
> of "glabel label" for generic labels and "tunefs -L" for UFS
> labels very well.
>
>
>
>   
>> note there I have also used glabel on the swap (command used was glabel
>> label /dev/ad10p1)
>>     
>
> A really honest question: What does the "p" in "ad10p1"
> indicate? I always thought swap partitions are something
> like "ad10b" (an own partition right after the root
> partition a).
>
>
>   
I partitioned my disk as GPT using gpart. no real reason, just because i
could ;)
this give partitions (p) instead of slices (s) (I think, not certain
thats what the p and s stand for but it makes sense)
I could also have used the gpt labels instead of generic glabel labels.
[root at ostracod /scratch]# dd if=/dev/zero of=dsk bs=1m count=10
10+0 records in
10+0 records out
10485760 bytes transferred in 0.055245 secs (189804954 bytes/sec)
[root at ostracod /scratch]# mdconfig -a -t vnode -f dsk -u 10
[root at ostracod /scratch]# gpart add -l MDDISK -t freebsd-ufs md10
md10p1 added
[root at ostracod /scratch]# ls /dev/gpt/
MDDISK

>   
>> One thing to note with label, if you mount/use the device by is raw
>> node, the label disapears.
>> [...]
>> This used to confuse me greatly :)
>>     
>
> Why make a label available for something to mount that is
> already mounted and cannot be accessed through this label
> while being mounted? :-)
>
> The kernel messages show such messages about removing labels
> as soon as devices are mounted in the "traditional" way.
>
>
>   
Indeed, that makes sense.

Vince
>
>