Article ID: 118837, created on Nov 24, 2013, last review on May 11, 2014

  • Applies to:
  • Virtuozzo
  • Virtuozzo hypervisor


A disk resizing operation on a virtual machine gets stuck in an infinite loop, the process sits in R state with 100% CPU usage:

[root@pcs ~]# ps aux | grep prl_disk
root      265065  100.0  0.0 387444 14560 ?        Rl   10:48   0:01 /usr/sbin/prl_disk_tool resize --hdd /vz/parallels/test.pvm/harddisk.hdd --size 317440M --resize_partition --comm /tmp/prl_shared_mem_{22fb6204-1a76-4af2-8218-f055c34a015f}1384422519

The process stack points to some operation with snumbd module:

[root@pcs ~]# cat /proc/98992/stack
[<ffffffff810a3361>] down+0x41/0x50
[<ffffffffa02b4758>] snumbd_make_request+0x68/0x380 [snumbd26]
[<ffffffff8126589e>] generic_make_request+0x24e/0x500
[<ffffffff81265c43>] submit_bio+0xf3/0x1a0
[<ffffffff811dc3d6>] submit_bh+0xf6/0x150
[<ffffffff811dec74>] block_read_full_page+0x264/0x3c0
[<ffffffff811e3ce8>] blkdev_readpage+0x18/0x20
[<ffffffff811475ba>] __do_page_cache_readahead+0x21a/0x220
[<ffffffff8114795f>] ondemand_readahead+0xcf/0x240
[<ffffffff81147c46>] page_cache_sync_readahead+0x86/0xa0
[<ffffffff81132236>] generic_file_read_iter+0x476/0x680
[<ffffffff811324cb>] generic_file_aio_read+0x8b/0xa0
[<ffffffff811e3143>] blkdev_aio_read+0x53/0xc0
[<ffffffff811a533a>] do_sync_read+0xfa/0x140
[<ffffffff811a5c05>] vfs_read+0xb5/0x1a0
[<ffffffff811a5f32>] sys_pread64+0x82/0xa0
[<ffffffff8100b102>] system_call_fastpath+0x16/0x1b
[<ffffffffffffffff>] 0xffffffffffffffff

There's possibly another D-state process that invokes /dev/snumbd1d device:

[root@pcs ~]# ps aux | grep snum
root       99004  1.6  0.0   8316   568 ?        D<   22:08   0:14 /sbin/blkid -o udev -p /dev/snumbd1d


The cause of the problem is a Red Hat kernel issue.


  1. Add the following udev rule that will ignore the snumbd device:

    [root@pcs ~]# cat  /etc/udev/rules.d/10-snumbd26-parallels.rules
    KERNEL=="snumbd*", OPTIONS="ignore_device"
  2. Reload the rules:

    [root@pcs ~]# udevadm control --reload-rules
  3. Reboot the node if the access to the device is still blocked.

NOTE: The resolution steps imply that the snumbd1d device is unnecessary on the system, as Parallels products do not make use of the snumbd26 module. If the device is required for some specific reason, contact Parallels Support to clarify the issue further.

Search Words



infinite loop

a26b38f94253cdfbf1028d72cf3a498b 2897d76d56d2010f4e3a28f864d69223 0dd5b9380c7d4884d77587f3eb0fa8ef

Email subscription for changes to this article
Save as PDF