run.py fails on XEN because of losetup failure

[tgrabiec]

$ sudo ./scripts/run.py -nv
/dev/loop15427: No such file or directory
Parsing config from /tmp/tmpl2OepB
libxl: error: libxl_device.c:265:libxl__device_disk_set_backend: Disk vdev=hda failed to stat: /dev/loop15427: No such file or directory
libxl: error: libxl_dm.c:1467:kill_device_model: unable to find device model pid in /local/domain/18/image/device-model-pid
libxl: error: libxl.c:1421:libxl__destroy_domid: libxl__destroy_device_model failed for 18
loop: can't delete device /dev/loop15427: No such file or directory

Looks like losetup fails:

sudo strace losetup /dev/loop123 build/release/usr.img
...
open("build/release/usr.img", O_RDWR)   = 3
open("/dev/loop123", O_RDWR)            = -1 ENOENT (No such file or directory)

It moves on with this patch:

diff --git a/scripts/run.py b/scripts/run.py
index 1a4957f..0f24845 100755
--- a/scripts/run.py
+++ b/scripts/run.py
@@ -233,7 +233,7 @@ def start_osv_xen(options):
         "vcpus=%s" % (options.vcpus),
         "maxcpus=%s" % (options.vcpus),
         "name='osv-%d'" % (os.getpid()),
-        "disk=['/dev/loop%s,raw,hda,rw']" % os.getpid(),
+        "disk=['%s,qcow2,hda,rw']" % (options.image_file),
         "serial='pty'",
         "paused=0",
         "on_crash='preserve'"
@@ -253,7 +253,7 @@ def start_osv_xen(options):
         stty_save()

         #create a loop device backed by image file
-        subprocess.call(["losetup", "/dev/loop%s" % os.getpid(), options.image_file])
+        # subprocess.call(["losetup", "/dev/loop%s" % os.getpid(), options.image_file]
         # Launch qemu
         cmdline = ["xl", "create" ]
         if not options.detach:

[glommer]

On Mon, Jun 16, 2014 at 8:27 PM, Tomasz Grabiec notifications@github.com
wrote:

$ sudo ./scripts/run.py -nv
/dev/loop15427: No such file or directory
Parsing config from /tmp/tmpl2OepB
libxl: error: libxl_device.c:265:libxl__device_disk_set_backend: Disk vdev=hda failed to stat: /dev/loop15427: No such file or directory
libxl: error: libxl_dm.c:1467:kill_device_model: unable to find device model pid in /local/domain/18/image/device-model-pid
libxl: error: libxl.c:1421:libxl__destroy_domid: libxl__destroy_device_model failed for 18
loop: can't delete device /dev/loop15427: No such file or directory

Looks like losetup fails:

sudo strace losetup /dev/loop123 build/release/usr.img
...
open("build/release/usr.img", O_RDWR) = 3
open("/dev/loop123", O_RDWR) = -1 ENOENT (No such file or directory)

It moves on with this patch:

diff --git a/scripts/run.py b/scripts/run.py
index 1a4957f..0f24845 100755
--- a/scripts/run.py
+++ b/scripts/run.py
@@ -233,7 +233,7 @@ def start_osv_xen(options):
"vcpus=%s" % (options.vcpus),
"maxcpus=%s" % (options.vcpus),
"name='osv-%d'" % (os.getpid()),

•    "disk=['/dev/loop%s,raw,hda,rw']" % os.getpid(),
  

•    "disk=['%s,qcow2,hda,rw']" % (options.image_file),
   "serial='pty'",
   "paused=0",
   "on_crash='preserve'"
  

  @@ -253,7 +253,7 @@ def start_osv_xen(options):
  stty_save()

   #create a loop device backed by image file
  

•    subprocess.call(["losetup", "/dev/loop%s" % os.getpid(), options.image_file])
  

•    # subprocess.call(["losetup", "/dev/loop%s" % os.getpid(), options.image_file]
   # Launch qemu
  

can't we instead check the result of the call and adjust the xen file
accordingly ?

     cmdline = ["xl", "create" ]
     if not options.detach:

—
Reply to this email directly or view it on GitHub
#344.

[sthgrau]

I ran into a similar losetup issue. I ended up having to compile a new version of losetup. The version I had (for LinuxMint 17) was 2.20.1-5.1ubuntu20.7. I compiled 2.27 (from ftp://ftp.kernel.org/pub/linux/utils/util-linux/v2.27/util-linux-2.27.tar.gz) and got past this problem.