INSTALL hint: create device btrfs-control

Hallo,

you should please add a hint in

.../git/mason/btrfs-progs-unstable.git

in the file "INSTALL" with at least the contents (please excuse my gerlish)

"btrfs" needs the device "btrfs-control". Perhaps you have to create it:

mknod /dev/btrfs-control c 10 234

Viele Gruesse!
Helmut

This should probably go in the problems FAQ [1] on the wiki
instead, since (a) it's really a problem with your distribution, and
(b) probably quite rare now that the majority of people are running distributions with udev.

What are the symptoms of not having the device node? (i.e. how
would I diagnose this particular problem?)

Hugo.

=== Hugo Mills: hugo@... carfax.org.uk | darksatanic.net | lug.org.uk ===
  PGP key: 515C238D from wwwkeys.eu.pgp.net or http://www.carfax.org.uk
          --- What part of "gestalt" don't you understand? ---           

Hallo, Hugo,

Du meintest am 04.08.11:

Sorry - it's no problem of my distribution; slackware uses udev.

But I run/maintain a very special system, without udev, with most modules integrated in the kernel etc. - it's for a special kind of servers, not for end user workstations.

And I believe btrfs should run not only for "the majority of people" without problems but also for systems which aren't "main stream".

Please wait some hours (or days); I'll have to build a test system (I won't change my running systems ...).

Viele Gruesse!
Helmut

Hallo, Hugo,

Du meintest am 04.08.11:

The days get shorter and shorter ...

running

btrfs device scan

produces something like

...
failed to read /dev/hdc
failed to read /dev/hdc8
failed to read /dev/lmscd
failed to read /dev/sdc1
failed to read /dev/sdb4
failed to open /dev/btrfs-control skipping device registration failed to read /dev/sdd8
failed to read /dev/hdc9
failed to read /dev/sdf8
...

a lot of error messages for not existing devices and 1 error messages on behalf of "btrfs-control".

My "/dev" directory has worked (without "udev") without any problem for more than 10 years, it contains (has contained) a lot of entries for nearly every case of disk. Therefore i had started with

btrfs device scan 2>/dev/null

and that deleted the "btrfs-control" error message too. And

btrfs device scan 2>&1 | grep -v 'failed to'

didn't help too.

btrfs-show

doesn't need "/dev/btrfs-control", it shows the existing btrfs partition(s).

If I remember correct: mounting works, writing works, reading works. And adding another partition works.

But there are strange messages when I try to resize or delete.

I'll try to reproduce these messages ...

Viele Gruesse!
Helmut

OK, that's the symptom we can put in the problems list for this
particular issue.

Goffredo published a patch to scan only the contents of
/proc/partitions, which should at least shut up the scan over the issue of non-existent devices in /dev. That's in the next integration-* branch (going out shortly... I'm just integrating and updating the balance management patches before i release).

Thanks. We can add those as symptoms to the solution later.

How's this?

https://btrfs.wiki.kernel.org/index.php/Problem_FAQ#I_get_the_message_.22failed_to_open_.2Fdev.2Fbtrfs-control_skipping_device_registration.22_from_.22btrfs_dev_scan.22

Hugo.

=== Hugo Mills: hugo@... carfax.org.uk | darksatanic.net | lug.org.uk ===
  PGP key: 515C238D from wwwkeys.eu.pgp.net or http://www.carfax.org.uk
       --- "I lost my leg in 1942.  Some bastard stole it in a ---       
                            pub in Pimlico."                             

Hallo, Hugo,

Du meintest am 04.08.11:

[...]

That sounds fine - thanks to you and to Goffredo!

I've just run the first test - finding reliable (for only this case) messages is a bad job ...

There's some problem in the Wiki. This new paragraph (?) is not shown when I call the "Problem_FAQ", it ends with the "Copy-on-write doesn't work" question.

When I look into the source code of that page I can see and read the paragraph, and it looks fine!

Viele Gruesse!
Helmut

Hi, Helmut,

Well, it's probably OK just to use this particular symptom for
now. You've generally got to run btrfs dev scan before you run anything else, so that message should turn up. I wouldn't spend ages hunting down additional failure messages.

Yes, I think there is a problem with it at the moment. The usual
one that happens every so often. I don't know what's up, but it seems to be something to do with the MySQL back-end.

At least the text looks good to you, so we'll leave it at that for
now, and it should be better the next time the kernel.org admins sort it out.

Hugo.

=== Hugo Mills: hugo@... carfax.org.uk | darksatanic.net | lug.org.uk ===
  PGP key: 515C238D from wwwkeys.eu.pgp.net or http://www.carfax.org.uk
  --- "What's so bad about being drunk?" "You ask a glass of water" ---  

Hallo, Hugo,

Du meintest am 04.08.11:

My usual way: "btrfs device scan" is part of a script under "/etc/ init.d", it's run while booting the machine, and even when I watch the machine while booting this special message isn't shown a long time. I'm searching ...

Viele Gruesse!
Helmut

Hallo, Hugo,

Du meintest am 04.08.11:

Maybe I've found the place of the (unique?) problem.

Partitions:

        sdb1    2 GByte
        sdb2    3 GByte
        sdb3    4 Gbyte

mkfs.btrfs -L Probe -m raid0 -d raid1 /dev/sdb1 mount LABEL=Probe /mnt/Probe
filling with about 600 MByte (1 *.iso)

btrfs device add /dev/sdb2 /mnt/Probe
btrfs filesystem balance /mnt/Probe

     (ok)
   btrfs device add /dev/sdb3 /mnt/Probe
     filling with a DVD *.iso (more than 4 GByte)
   btrfs filesystem balance /mnt/Probe
     (ok)

And then:

btrfs device delete /dev/sdb2 /mnt/Probe

works fine, if /dev/btrfs-control exists

crashes (shown with "dmesg") if this device doesn't exist.

Viele Gruesse!
Helmut