Buffer I/O error, with latest losetup on first run

Buffer I/O error, with latest losetup on first run

[Gerardo Exequiel Pozzi]

Hello

Using (losetup from util-linux 2.20.620-4360-dirty) and Linux-3.2.1.

Using losetup as first time, if loop module is not loaded, losetup -f 
file-equal-or-below-1MiB I get: "Buffer I/O error on device loop0, 
logical block 0" (sometimes get one message, or two or three, ...)

Can be reproducible again if loop module is unloaded then execute losetup.
Does not happen if module is loaded previously to losetup.
Happens independently if /dev/loop0 exists or not.

Does not happen with losetup from 2.20 under the same conditions.

Maybe a race-condition?

-- 
Gerardo Exequiel Pozzi
\cos^2\alpha + \sin^2\alpha = 1

Re: Buffer I/O error, with latest losetup on first run

[Karel Zak]

On Tue, Jan 17, 2012 at 10:59:59PM -0300, Gerardo Exequiel Pozzi wrote:
> Using (losetup from util-linux 2.20.620-4360-dirty) and Linux-3.2.1.
> 
> Using losetup as first time, if loop module is not loaded, losetup -f
> file-equal-or-below-1MiB I get: "Buffer I/O error on device loop0, logical
> block 0" (sometimes get one message, or two or three, ...)

It does not sound like userspace problem, I guess that the message comes
from kernel (dmesg). It would be also nice to have strace output.

> 
> Can be reproducible again if loop module is unloaded then execute losetup.
> Does not happen if module is loaded previously to losetup.
> Happens independently if /dev/loop0 exists or not.
> 
> Does not happen with losetup from 2.20 under the same conditions.

The new losetup uses /dev/loop-control to get the new free loop device.

Kay, any idea?

    Karel

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

Re: Buffer I/O error, with latest losetup on first run

[Kay Sievers]

On Wed, Jan 18, 2012 at 13:13, Karel Zak <kzak@redhat.com> wrote:
> On Tue, Jan 17, 2012 at 10:59:59PM -0300, Gerardo Exequiel Pozzi wrote:
>> Using (losetup from util-linux 2.20.620-4360-dirty) and Linux-3.2.1.
>>
>> Using losetup as first time, if loop module is not loaded, losetup -f
>> file-equal-or-below-1MiB I get: "Buffer I/O error on device loop0, logical
>> block 0" (sometimes get one message, or two or three, ...)
>
> It does not sound like userspace problem, I guess that the message comes
> from kernel (dmesg). It would be also nice to have strace output.
>
>>
>> Can be reproducible again if loop module is unloaded then execute losetup.
>> Does not happen if module is loaded previously to losetup.
>> Happens independently if /dev/loop0 exists or not.
>>
>> Does not happen with losetup from 2.20 under the same conditions.
>
> The new losetup uses /dev/loop-control to get the new free loop device.
>
> Kay, any idea?

No really. Accessing /dev/loop-control should block losetup, and
synchronously load the look.ko module, before doing any further
operation in losetup.

Kay

Re: Buffer I/O error, with latest losetup on first run

[Gerardo Exequiel Pozzi]

On 01/18/2012 09:13 AM, Karel Zak wrote:
> On Tue, Jan 17, 2012 at 10:59:59PM -0300, Gerardo Exequiel Pozzi wrote:
>> Using (losetup from util-linux 2.20.620-4360-dirty) and Linux-3.2.1.
>>
>> Using losetup as first time, if loop module is not loaded, losetup -f
>> file-equal-or-below-1MiB I get: "Buffer I/O error on device loop0, logical
>> block 0" (sometimes get one message, or two or three, ...)
> It does not sound like userspace problem, I guess that the message comes
> from kernel (dmesg). It would be also nice to have strace output.
Yes, message is from kernel. Remember that sometimes works without any 
error, its random.

# strace losetup -f 1M
execve("/sbin/losetup", ["losetup", "-f", "1M"], [/* 13 vars */]) = 0
brk(0)                                  = 0x8ede000
mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 
0) = 0xb7845000
access("/etc/ld.so.preload", R_OK)      = -1 ENOENT (No such file or 
directory)
open("/etc/ld.so.cache", O_RDONLY|O_CLOEXEC) = 3
fstat64(3, {st_mode=S_IFREG|0644, st_size=26081, ...}) = 0
mmap2(NULL, 26081, PROT_READ, MAP_PRIVATE, 3, 0) = 0xb783e000
close(3)                                = 0
open("/lib/libc.so.6", O_RDONLY|O_CLOEXEC) = 3
read(3, 
"\177ELF\1\1\1\3\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0\340\225\1\0004\0\0\0"..., 
512) = 512
fstat64(3, {st_mode=S_IFREG|0755, st_size=1954497, ...}) = 0
mmap2(NULL, 1710788, PROT_READ|PROT_EXEC, MAP_PRIVATE|MAP_DENYWRITE, 3, 
0) = 0xb769c000
mmap2(0xb7838000, 12288, PROT_READ|PROT_WRITE, 
MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x19c) = 0xb7838000
mmap2(0xb783b000, 10948, PROT_READ|PROT_WRITE, 
MAP_PRIVATE|MAP_FIXED|MAP_ANONYMOUS, -1, 0) = 0xb783b000
close(3)                                = 0
mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 
0) = 0xb769b000
set_thread_area({entry_number:-1 -> 6, base_addr:0xb769b6c0, 
limit:1048575, seg_32bit:1, contents:0, read_exec_only:0, 
limit_in_pages:1, seg_not_present:0, useable:1}) = 0
mprotect(0xb7838000, 8192, PROT_READ)   = 0
mprotect(0x8052000, 4096, PROT_READ)    = 0
mprotect(0xb7867000, 4096, PROT_READ)   = 0
munmap(0xb783e000, 26081)               = 0
brk(0)                                  = 0x8ede000
brk(0x8eff000)                          = 0x8eff000
open("/usr/lib/locale/locale-archive", O_RDONLY|O_LARGEFILE|O_CLOEXEC) = 3
fstat64(3, {st_mode=S_IFREG|0644, st_size=3111552, ...}) = 0
mmap2(NULL, 2097152, PROT_READ, MAP_PRIVATE, 3, 0) = 0xb749b000
close(3)                                = 0
close(0)                                = 0
close(0)                                = -1 EBADF (Bad file descriptor)
uname({sys="Linux", node="archsrv", ...}) = 0
stat64("/dev/loop-control", {st_mode=S_IFCHR, st_rdev=makedev(10, 237), 
...}) = 0
open("/dev/loop-control", O_RDWR|O_LARGEFILE) = 0
ioctl(0, 0x4c82, 0xb783bda0)            = 0
close(0)                                = 0
getcwd("/root", 4095)                   = 6
readlink("/root/1M", 0xbf9859ab, 4096)  = -1 EINVAL (Invalid argument)
open("/root/1M", O_RDWR|O_LARGEFILE)    = 0
open("/dev/loop0", O_RDWR|O_LARGEFILE)  = 3
ioctl(3, LOOP_SET_FD[   32.948747] Buffer I/O error on device loop0, 
logical block 0
[   32.949764] Buffer I/O error on device loop0, logical block 0
, 0)                = 0
close(0)                                = 0
ioctl(3, LOOP_SET_STATUS64, 0xbf986b64) = 0
close(3)                                = 0
exit_group(0)                           = ?


>> Can be reproducible again if loop module is unloaded then execute losetup.
>> Does not happen if module is loaded previously to losetup.
>> Happens independently if /dev/loop0 exists or not.
>>
>> Does not happen with losetup from 2.20 under the same conditions.
> The new losetup uses /dev/loop-control to get the new free loop device.

Yes, I know, I expected something like this for a long time. Many thanks :)

I more info is needed please let me know.

-- 
Gerardo Exequiel Pozzi
\cos^2\alpha + \sin^2\alpha = 1