Fwd: dmcache development

111 views
Skip to first unread message

Ming Zhao

unread,
Sep 2, 2011, 9:56:29 AM9/2/11
to dm-c...@googlegroups.com
The set-up that I can consistently crash:

       Xen 4.0.1 or Xen 4.1.1 (gitco.de builds) booted with dom0_mem=1024M
       Kernel 2.6.18 or kernel 3.0 booted with pci=nomsi
       iSCSI-link to a remote machine running 2.6.18
       The iSCSI host exports a scsi device that the dom0 treats as an LVM PV.
       The PV contains a VG with one LV per partition on each guest.

       For the test-setup, I created a 128MB ramdisk (ramdisk_size=131072)
       I used dmc-setup to bind the 10G lvm partition containing a
guest's root partition and /dev/ram0:

       # ./dmc-setup.pl -o /dev/ivpsvg-001/tvps054root -c /dev/ram0 -n
tvps054cache -s 16384

       Mounting the mapped device worked.
       Reading and writing to the mapped device worked.

       Next I changed the guest's configuration to use the mapped device:

               # Generated by zero1 21093
               name = "tvps054"

               memory = 128
               maxmem = 256
               disk = [ 'phy:/dev/mapper/tvps054cache,xvda1,w',
'phy:/dev/mapper/ivpsvg--001-tvps054swap,xvda2,w' ]
               vif = [ 'mac=00:16:3e:7d:6e:ba, bridge=xenbr300'  ]
               uuid = "7d6ebadc-4353-4c9e-828f-c8d5163360fb"
               kernel = "/boot/vmlinuz-2.6.32.8-xls-xenu"
               # ramdisk = ""
               root = "/dev/xvda1 ro"
               stdvga=0
               vnc=0
               apic=1
               pae=1
               vcpus = 1
               on_reboot = 'destroy'
               on_crash = 'restart'

       It's the same configuration as the working one, with xvda1
being the only change.

       After an 'xm create -c' of the domain, the domU kernel
initializes normally, up to:

               NET: Registered protocol family 15
               registered taskstats version 1
               XENBUS: Device with no driver: device/console/0

       Then the dom0 crashes, presumably because the domU starts
accessing the disk device.

I haven't set up a debugger yet, but I attached an old school
photograph of the kernel panic last time I managed to catch it.

Cheers,
Pim

IMG_0964.JPG
Reply all
Reply to author
Forward
0 new messages