If the normal fs tree is hosed and the user has multiple subvolumes it's handy
to be able to specify just one of the subvolumes to restore.  It's also handy if
a user only wants to restore say /home instead of his entire disk.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>

This will make the restore program fall back on other mirrors if it fails to
decompress an extent for whatever reason.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>

If we hit a bad disk and the read doesn't work, try other mirrors in case we
have other disks with good copies.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>

… mirrors exist

Fixes for my Raid10 restore issues.

Signed-off-by: David Marcin <djmarcin@google.com>

btrfs-progs: Check metadata mirrors in find-root.

…ameters

search_dir() recurses down the btrfs tree, and used to take the output
path for every item (i.e. in the running system, output root directory
concatenated with btrfs-local pathname) passed as the only path
parameter. Moving the output root directory to a separate parameter
and passing the btrfs-local pathname for each file and directory
separately allows easy filtering based on the btrfs-local pathname.

Signed-off-by: Peter Stuge <peter@stuge.se>
Signed-off-by: Josef Bacik <josef@redhat.com>

The option -m is used to specify the regex string. -c is used to
specify case insensitive matching. -i was already taken.

In order to restore only a single folder somewhere in the btrfs
tree, it is unfortunately neccessary to construct a slightly
nontrivial regex, e.g.:

restore -m '^/(|home(|/username(|/Desktop(|/.*))))$' /dev/sdb2 /output

This is needed in order to match each directory along the way to the
Desktop directory, as well as all contents below the Desktop directory.

Signed-off-by: Peter Stuge <peter@stuge.se>
Signed-off-by: Josef Bacik <josef@redhat.com>

Signed-off-by: David Marcin <djmarcin@google.com>

Signed-off-by: David Marcin <djmarcin@google.com>

…ning the ctree.

Signed-off-by: David Marcin <djmarcin@google.com>

…nfirmation to btrfs-select-super.

Signed-off-by: David Marcin <djmarcin@google.com>

btrfs-progs: In find-root, dump bytenr for every slot.

btrfs-progs: Utilities for dumping and rolling back to oldest super on filesystem

This patch simply adds support to decompress lzo compressed extents in restore.

Signed-off-by: Josef Bacik <josef@redhat.com>

We were unconditionally executing our regular expression, even though we may not
have one, so check to make sure mreg is not null before calling regexec.
Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>

…ced by refactoring.

Signed-off-by: David Marcin <djmarcin@google.com>

btrfs-progs: Fix compilation errors with btrfs-select-super.c introduced...

We don't need most of the roots when doing restore, like the extent tree.  So if
the recovery open_ctree fails because it's trying to open one of these useless
roots just fall back to open_ctree_broken.  This will just open the chunk root
which is the bare minimum of what we need to operate.  Then from there we can
setup the tree_root and the fs_root, and if either of those are gone we can't
restore anyway.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>

Return an error instead of BUG()'ing out.

Signed-off-by: Josef Bacik <josef@redhat.com>

Don't BUG(), return an error so the recovery program can work its mojo.

Signed-off-by: Josef Bacik <josef@redhat.com>

We were checking for a null root coming back from btrfs_read_fs_root but thats
not right since it returns a ERR_PTR.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>

Since restore has the ability to open really really screwed up file systems, add
a list roots option to it so we can still get the contents of the tree root on a
horribly broken fs.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>

In order to figure out what exactly is broken on a fs we need to spit out the
current offset we are on and the size of the fs to know if the super is wrong
and we just need to ignore it, or if the offset we got is bad and we should just
keep searching.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>

Trying to track down why we can't find roots, add some verbose output so we know
what chunks we're scanning and when we move to new chunks.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>

We were using the wrong variable for the root location if we specified -f when
doing restore.  Fix this.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>

Our logical offsets may be beyond what we think the size of the disk is, so our
check is bogus, remove it.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>

If we have to build our fs_info by hand don't read from the cache when looking
for the fs_root just in case we set something up last time.  Also actually print
the right error, not the root which is ok.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>

We need to know if we find a valid fs tree when doing find root, so print the
objectid of the roots we find when we find a tree root.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>

We need to be able to handle the case where we want to restore from a specific
root if the fs is really really really toast, this patch does that.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>

It may be used elsewhere and in the case of a broken fs it won't be there at all
and it makes an assertion trip.  Thanks,

Signed-off-by: Josef Bacik <josef@redhat.com>

Small fixes to compilation/linking

fix for a simple typo