[Libguestfs] [PATCH 0/2] error accessing harddrive partitions

[Pavel Butsykin]

virt-v2v fails when traverses each partition:

inspect_sources.ml:
...
  and is_uefi_bootable_part part =
    let dev = g#part_to_dev part in
    parttype_is_gpt dev && is_uefi_ESP dev part
  in

  let partitions = Array.to_list (g#list_partitions ()) in
  let partitions = List.filter is_uefi_bootable_part partitions in

log:
...
libguestfs: trace: v2v: list_partitions
guestfsd: main_loop: new request, len 0x28
guestfsd: main_loop: proc 8 (list_partitions) took 0.00 seconds
libguestfs: trace: v2v: list_partitions = ["/dev/sda1", "/dev/sda2", "/dev/sda3",
"/dev/sdb1", "/dev/sdb2", "/dev/sdb5", "/dev/sdc1", "/dev/sdc2", "/dev/sdc5", 
"/dev/sdd1", "/dev/sdd2", "/dev/sdd5", "/dev/sde1", "/dev/sde2", "/dev/sde5", 
"/dev/sdf1", "/dev/sdf2", "/dev/sdf5", "/dev/sdg1", "/dev/sdg2", "/dev/sdg5"]
...
libguestfs: trace: v2v: part_get_parttype "/dev/sdb"
guestfsd: main_loop: new request, len 0x34
commandrvf: stdout=y stderr=y flags=0x0
commandrvf: parted -m -s -- /dev/sdb unit b print
libguestfs: trace: v2v: part_get_parttype = "msdos"
libguestfs: trace: v2v: part_get_parttype "/dev/sdb"
guestfsd: main_loop: proc 214 (part_get_parttype) took 0.00 seconds
guestfsd: main_loop: new request, len 0x34
commandrvf: stdout=y stderr=y flags=0x0
commandrvf: parted -m -s -- /dev/sdb unit b print
libguestfs: trace: v2v: part_get_parttype = "msdos"
libguestfs: trace: v2v: part_to_dev "/dev/sdb5"
guestfsd: main_loop: proc 214 (part_get_parttype) took 0.00 seconds
[    9.164666]  sdb: sdb1 sdb2 < sdb5 >
guestfsd: main_loop: new request, len 0x38
/dev/sdb5: No such file or directory

call part_to_dev "/dev/sdb5" failed in is_root_device():

part_to_dev_stub()
RESOLVE_DEVICE()
is_root_device ():
{
  struct stat statbuf;
  if (stat (device, &statbuf) == -1) {
    perror (device);
    return 0;
...


let's look at the udev event log after the error:

KERNEL[10.527855] change   /devices/pci0000:00/0000:00:03.0/virtio1/host2/target2:0:1/2:0:1:0/block/sdb (block)
UDEV  [10.528693] change   /devices/pci0000:00/0000:00:03.0/virtio1/host2/target2:0:0/2:0:0:0/block/sda/sda1 (block)
KERNEL[10.529156] add      /devices/pci0000:00/0000:00:03.0/virtio1/host2/target2:0:1/2:0:1:0/block/sdb/sdb1 (block)
KERNEL[10.529618] add      /devices/pci0000:00/0000:00:03.0/virtio1/host2/target2:0:1/2:0:1:0/block/sdb/sdb2 (block)
KERNEL[10.530096] add      /devices/pci0000:00/0000:00:03.0/virtio1/host2/target2:0:1/2:0:1:0/block/sdb/sdb5 (block)
...

Apparently is_root_device("/dev/sdb5") was called before "/dev/sdb5" was added.
To solve issue we need to wait adding a partition before we can access it.
Another solution if we wait for adding all partitions somewhere in one place.
But there is at least one drawback, not for each task need to touch the disk
partitions and waiting for all partitions will slow such tasks.


Pavel Butsykin (2):
  daemon: run 'udevadm settle' with --exit-if-exists option.
  daemon: add udev_settle_file to is_root_device

 daemon/daemon.h   |  2 ++
 daemon/guestfsd.c | 23 +++++++++++++++++++----
 2 files changed, 21 insertions(+), 4 deletions(-)

-- 
2.11.0