Skip to content
modified mdb for examining zfs on disk
Latest commit 2fcbf53 Aug 12, 2013 Super-User add link to blog post in
Failed to load latest commit information.
mdb first time with binaries Aug 9, 2013
zuncompress.c first time with binaries Aug 9, 2013


modified mdb for examining zfs on disk

This contains changes for usr/src/cmd/mdb/common/mdb/mdb_rawfile.c, usr/src/cmd/mdb/common/modules/zfs/zfs.c, and source for zuncompress.c and rawzfs.c Also included are binaries:

 mdb - with changes to look at ZFS on disk - slight change to be used with mdb - new mdb dmod for the ::walk uberblock command
 zuncompress - uncompress data

To see these in use, see the Bruning Questions blog post at

To build, start with and follow all of the instructions there. Then copy mdb_rawfile.c from here to smartos-live/projects/illumos/usr/src/cmd/mdb/common/mdb/mdb_rawfile.c, zfs.c from here to smartos-live/projects/illumos/usr/src/cmd/mdb/common/modules/zfs.c, and rawzfs.c to smartos-live/projects/illumos/usr/src/cmd/mdb/common/modules/rawzfs.c. You may need to edit the mdb Makefile to get to build, but it should be pretty straight forward. To build zuncompress: gcc -m64 -I ./smartos-live/projects/illumos/usr/src/uts/common/fs/zfs -lzfs -lzpool -o zuncompress zuncompress.c

This is very much a work in progress. In its final form, I expect someone will make a ZFS on disk target for mdb, and the hacks in mdb_rawfile.c can go away. There are lots of dcmds and walkers that could be written that would be interesting. For instance, object_id_#::dnodefind [-d dataset] [-p zpool]. But the best would be: disk_address::zprint [-c compression_algorithm] datatype Then there would be no need to use zuncompress, and you could do everything in 1 mdb session.

Something went wrong with that request. Please try again.