scrub: Inappropriate ioctl for device

Hallo,

I'm just playing with "btrfs scrub".

Kernel 3.1 (self made)
btrfs integration-20111030 (Hugo Mills)

I have a bundle of 3 2-TByte-disks (data raid0, metadata raid1).

Since some (few) weeks one of the disks makes errors (I've reported the problems in this mailing list).

The bundle uses the devices /dev/sdb1, /dev/sdc1 and /dev/sdg1, "btrfs filesystem show" shows these informations.

Mounting doesn't work immediately; most times I have to run the "mount" command three times - strange.

When I run

btrfs scrub start -r /dev/sdb1

(or another of the three disks) then I only get the error message

ERROR: getting dev info for scrub failed: Inappropriate ioctl for
device

I don't use "udev", "ls -l /dev/sdb1" shows

brw-r----- 1 root disk 8, 17 29. Apr 1995 /dev/sdb1

Running the command via "strace" shows the last lines

set_thread_area({entry_number:-1 -> 6, base_addr:0x401b6b80, limit:1048575, seg_32bit:1, contents:0, read_exec_only:0, limit_in_pages:1, seg_not_present:0, useable:1}) = 0

mprotect(0x401b0000, 8192, PROT_READ)   = 0
mprotect(0x4004c000, 4096, PROT_READ)   = 0
mprotect(0x4001d000, 4096, PROT_READ)   = 0
munmap(0x40021000, 89277)               = 0
set_tid_address(0x401b6be8)             = 10400
set_robust_list(0x401b6bf0, 0xc)        = 0

futex(0xbffe4c00, FUTEX_WAKE_PRIVATE, 1) = 0
futex(0xbffe4c00, FUTEX_WAIT_BITSET_PRIVATE|FUTEX_CLOCK_REALTIME, 1, NULL, bffe4c10) = -1 EAGAIN (Resource temporarily unavailable) rt_sigaction(SIGRTMIN, {0x4003b520, [], SA_SIGINFO}, NULL, 8) = 0
rt_sigaction(SIGRT_1, {0x4003b5a0, [], SA_RESTART|SA_SIGINFO}, NULL, 8) = 0
rt_sigprocmask(SIG_UNBLOCK, [RTMIN RT_1], NULL, 8) = 0
getrlimit(RLIMIT_STACK, {rlim_cur=RLIM_INFINITY, rlim_max=RLIM_INFINITY}) = 0
uname({sys="Linux", node="Arktur.wm8.hullen.de", ...}) = 0

brk(0)                                  = 0x806d000
brk(0x808e000)                          = 0x808e000
mkdir("/var", 0777)                     = -1 EEXIST (File exists)
mkdir("/var/lib", 0777)                 = -1 EEXIST (File exists)
mkdir("/var/lib/btrfs", 0777)           = -1 EEXIST (File exists)

stat64("/dev/sdb1", {st_mode=S_IFBLK|0640, st_rdev=makedev(8, 17), ...}) = 0
open("/dev/sdb1", O_RDWR|O_LARGEFILE) = 3
ioctl(3, 0x8400941f, 0xbffe4554) = -1 ENOTTY (Inappropriate ioctl for device)
write(2, "ERROR: getting dev info for scru"..., 73ERROR: getting dev info for scrub failed: Inappropriate ioctl for device ) = 73
close(3) = 0
exit_group(1) = ?

--------------------------------------------------

Changing the rights of "/dev/sdb1" to 660 doesn't help - same nasty behaviour.

What goes wrong?

Viele Gruesse!
Helmut

To be able to run mount only once you have to run 'btrfs dev scan' on startup (after each reboot or just before mounting). If that's not an option you can use -o devices mount option.

Scrubber runs on a mounted filesystem.

Thanks,

Ilya

Hallo, Ilya,

Du meintest am 01.11.11:

That command has run, without any error.

There seems to be another problem.

"df" tells that the bundle of disks is mounted - there too seems to be another problem.

But

btrfs scrub start -r /path/to/mountpoint

works - is there a mistake in the description of the command?

In about 16 hours I'll know more ...

Viele Gruesse!
Helmut

I observed failed mounts between 2 consecutive xfstests (random two, no pattern) while the umount/mount went just fine when I tried it manually. IIRC it was in 3.1-rc9 and one of Josef's branches, but I haven't seen it since and was not able to reproduce it reliably.

I think it's not a problem from the userspace side (like access rights) but some unfinished device cleanup from the previous umount which blocks new mount.

david