platform/upstream/btrfs-progs.git
7 years agobtrfs-progs: test: fix run_check_stdout() call _fail()
Misono, Tomohiro [Tue, 5 Sep 2017 05:49:02 +0000 (14:49 +0900)]
btrfs-progs: test: fix run_check_stdout() call _fail()

run_check_stdout() uses "... | tee ... || _fail".  However, since tee
won't fail, _fail() is not called even if first command fails.

Fix this by checking PIPESTATUS in the end.

Signed-off-by: Tomohiro Misono <misono.tomohiro@jp.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: Add test case for mkfs --rootdir parameter
Qu Wenruo [Tue, 5 Sep 2017 01:24:15 +0000 (10:24 +0900)]
btrfs-progs: tests: Add test case for mkfs --rootdir parameter

Add test case which checks if -r|--rootdir mkfs option can handle
symlink/char/block/fifo files.

Signed-off-by: Qu Wenruo <quwenruo.btrfs@gmx.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: mkfs: Fix wrong file type for dir items and indexes when specifying...
Qu Wenruo [Mon, 4 Sep 2017 03:43:19 +0000 (12:43 +0900)]
btrfs-progs: mkfs: Fix wrong file type for dir items and indexes when specifying root directory

[Bug]
If using mkfs.btrfs with "-r" parameter and specified directory has
fifo/socket/char/block special file, then created filesystem can't pass
fsck:

------
checking fs roots
unresolved ref dir 241158 index 3 namelen 9 name S.dirmngr filetype 0 errors 80, filetype mismatch
ERROR: errors found in fs roots
------

[Reason]
Btrfs dir items/indexes records inode type, while "-r" only handles
directories, regular files and symlink, it makes such special files type
to be regular file and caused the problem.

[Fix]
Add missing types for add_directory_items(), so that result of
"mkfs.btrfs -r" can pass mkfs.

Signed-off-by: Qu Wenruo <quwenruo.btrfs@gmx.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: mkfs: Move the tree root creation to own function
Gu Jinxiang [Fri, 25 Aug 2017 07:23:36 +0000 (15:23 +0800)]
btrfs-progs: mkfs: Move the tree root creation to own function

make_btrfs is too long to understand, make creatation of root tree
in a function.

Some of the tree roots are now created in a loop, where the code is just
copypasted. We now make use of the reference_root_table to translate
block index to root objectid.

Signed-off-by: Gu Jinxiang <gujx@cn.fujitsu.com>
[ updated changelog ]
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: missing device and slack space report
David Sterba [Mon, 4 Sep 2017 16:33:48 +0000 (18:33 +0200)]
btrfs-progs: tests: missing device and slack space report

Verify that a missing device will not result in reporting a negative
value interpreted as 16EiB.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: device usage: don't calculate slack on missing device
Patrik Lundquist [Thu, 31 Aug 2017 11:00:24 +0000 (13:00 +0200)]
btrfs-progs: device usage: don't calculate slack on missing device

Print      Device slack:              0.00B
instead of Device slack:           16.00EiB

Signed-off-by: Patrik Lundquist <patrik.lundquist@gmail.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: inspect rootid: Allow a file to be specified
Misono, Tomohiro [Mon, 4 Sep 2017 05:05:34 +0000 (14:05 +0900)]
btrfs-progs: inspect rootid: Allow a file to be specified

Since cmd_inspect_rootid() calls btrfs_open_dir(), it rejects a file to
be specified. But as the document says, a file should be supported.

This patch introduces btrfs_open_file_or_dir(), which is a counterpart
of btrfs_open_dir(), to safely check and open btrfs file or directory.
The original btrfs_open_dir() content is moved to btrfs_open() and shared
by both function.

Signed-off-by: Tomohiro Misono <misono.tomohiro@jp.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: convert misc/011-delete-missing-device to loopdevs
David Sterba [Fri, 1 Sep 2017 18:18:20 +0000 (20:18 +0200)]
btrfs-progs: tests: convert misc/011-delete-missing-device to loopdevs

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: convert misc/006-image-on-missing-device to loopdevs
David Sterba [Fri, 1 Sep 2017 18:14:55 +0000 (20:14 +0200)]
btrfs-progs: tests: convert misc/006-image-on-missing-device to loopdevs

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: move loopdev helpers out of the testcase to common
David Sterba [Fri, 1 Sep 2017 18:14:03 +0000 (20:14 +0200)]
btrfs-progs: tests: move loopdev helpers out of the testcase to common

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: cleanup loop device helpers
David Sterba [Fri, 1 Sep 2017 17:32:16 +0000 (19:32 +0200)]
btrfs-progs: tests: cleanup loop device helpers

Make the loop device helpers a bit more generic before moving them to
the common helpers.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: print-tree: factor out extent_csum dump
David Sterba [Thu, 31 Aug 2017 15:07:00 +0000 (17:07 +0200)]
btrfs-progs: print-tree: factor out extent_csum dump

Factor out code to own helper and tweak the format so it matches the
rest.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: print the csum length in debug-tree
Josef Bacik [Fri, 25 Aug 2017 15:11:46 +0000 (11:11 -0400)]
btrfs-progs: print the csum length in debug-tree

While looking at a log of a corrupted fs I needed to verify we were
missing csums for a given range.  Make this easier by printing out the
range of bytes a csum item covers.

Signed-off-by: Josef Bacik <jbacik@fb.com>
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: add testcase for 'fi du' and empty subvol
David Sterba [Fri, 1 Sep 2017 14:45:34 +0000 (16:45 +0200)]
btrfs-progs: tests: add testcase for 'fi du' and empty subvol

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: fi du: don't call lookup_path_rootid for BTRFS_EMPTY_SUBVOL_DIR_OBJECTID
Goffredo Baroncelli [Fri, 18 Aug 2017 07:04:21 +0000 (09:04 +0200)]
btrfs-progs: fi du: don't call lookup_path_rootid for BTRFS_EMPTY_SUBVOL_DIR_OBJECTID

When ino is BTRFS_EMPTY_SUBVOL_DIR_OBJECTID, the item is not referred to
any file-tree. So lookup_path_rootid() doesn't return any meaningful
value.

As was reported, this can be triggered by

$ btrfs sub create test1
$ btrfs sub create test1/test2
$ btrfs sub snap test1 test1.snap
$ btrfs fi du -s test1
  Total   Exclusive  Set shared  Filename
  0.00B       0.00B       0.00B  test1
$ btrfs fi du -s test1.snap
  Total   Exclusive  Set shared  Filename
ERROR: cannot check space of 'test1.snap': Inappropriate ioctl for device

Signed-off-by: Goffredo Baroncelli <kreijack@inwind.it>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: reset the ret value when ignoring an error from du_add_file
Goffredo Baroncelli [Fri, 18 Aug 2017 07:04:20 +0000 (09:04 +0200)]
btrfs-progs: reset the ret value when ignoring an error from du_add_file

In du_walk_dir(), when du_add_file() returns an error it is usually
ignored. However if the error is returned querying the last item, the
error is returned to the caller.

Signed-off-by: Goffredo Baroncelli <kreijack@inwind.it>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: add test for check --force
David Sterba [Thu, 31 Aug 2017 16:49:38 +0000 (18:49 +0200)]
btrfs-progs: tests: add test for check --force

Basic test of the --force functionality, on an empty filesystem.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: add option to skip mount checks
David Sterba [Thu, 31 Aug 2017 16:34:37 +0000 (18:34 +0200)]
btrfs-progs: check: add option to skip mount checks

Sometimes it's needed to do a check on a mounted filesystem. This should
work fine on a quiescent filesystem or a read-only mount. Changes on the
block device done by kernel might confuse the userspace checker and it
might crash when it reads some stale data.

Repair without mount checks is not supported right now.

Signed-off-by: David Sterba <dsterba@suse.cz>
7 years agobtrfs-progs: print-tree: factor out temporary_item dump
David Sterba [Thu, 31 Aug 2017 15:07:00 +0000 (17:07 +0200)]
btrfs-progs: print-tree: factor out temporary_item dump

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: print-tree: factor out persistent_item dump
David Sterba [Thu, 31 Aug 2017 15:07:00 +0000 (17:07 +0200)]
btrfs-progs: print-tree: factor out persistent_item dump

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: print-tree: factor out qgroup_limit dump
David Sterba [Thu, 31 Aug 2017 15:07:00 +0000 (17:07 +0200)]
btrfs-progs: print-tree: factor out qgroup_limit dump

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: print-tree: factor out qgroup_info dump
David Sterba [Thu, 31 Aug 2017 15:07:00 +0000 (17:07 +0200)]
btrfs-progs: print-tree: factor out qgroup_info dump

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: print-tree: factor out qgroup_status dump
David Sterba [Thu, 31 Aug 2017 15:07:00 +0000 (17:07 +0200)]
btrfs-progs: print-tree: factor out qgroup_status dump

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: print-tree: factor out dev_extent dump
David Sterba [Thu, 31 Aug 2017 15:07:00 +0000 (17:07 +0200)]
btrfs-progs: print-tree: factor out dev_extent dump

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: print-tree: factor out free_space_info dump
David Sterba [Thu, 31 Aug 2017 15:07:00 +0000 (17:07 +0200)]
btrfs-progs: print-tree: factor out free_space_info dump

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: print-tree: factor out shared_data_ref dump
David Sterba [Thu, 31 Aug 2017 15:07:00 +0000 (17:07 +0200)]
btrfs-progs: print-tree: factor out shared_data_ref dump

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: print-tree: factor out extent_data_ref dump
David Sterba [Thu, 31 Aug 2017 15:07:00 +0000 (17:07 +0200)]
btrfs-progs: print-tree: factor out extent_data_ref dump

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: print-tree: factor out block_group_item dump
David Sterba [Thu, 31 Aug 2017 15:07:00 +0000 (17:07 +0200)]
btrfs-progs: print-tree: factor out block_group_item dump

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: print-tree: don't print the item type for all items
David Sterba [Thu, 31 Aug 2017 14:57:47 +0000 (16:57 +0200)]
btrfs-progs: print-tree: don't print the item type for all items

In some cases it's clear from the context which item is being printed,
so we can remove them. If the item has no data, some description is
still desired (eg. orphan or various backrefs).

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: print-tree: rename item callbacks to match the key name
David Sterba [Thu, 31 Aug 2017 14:34:15 +0000 (16:34 +0200)]
btrfs-progs: print-tree: rename item callbacks to match the key name

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: move fs roots check mode switch to a helper
David Sterba [Wed, 30 Aug 2017 17:47:50 +0000 (19:47 +0200)]
btrfs-progs: check: move fs roots check mode switch to a helper

Simplify main a bit.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: cleanup use of fs_info in check_fs_roots
David Sterba [Wed, 30 Aug 2017 17:52:44 +0000 (19:52 +0200)]
btrfs-progs: check: cleanup use of fs_info in check_fs_roots

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: pass fs_info to check_fs_roots
David Sterba [Wed, 30 Aug 2017 17:50:31 +0000 (19:50 +0200)]
btrfs-progs: check: pass fs_info to check_fs_roots

The root pointer is not used anyway, will be cleaned up next.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: move chunk and extent check mode switch to a helper
David Sterba [Wed, 30 Aug 2017 17:47:50 +0000 (19:47 +0200)]
btrfs-progs: check: move chunk and extent check mode switch to a helper

Simplify main a bit.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: pass fs_info to check_chunks_and_extents_v2
David Sterba [Wed, 30 Aug 2017 17:38:02 +0000 (19:38 +0200)]
btrfs-progs: check: pass fs_info to check_chunks_and_extents_v2

The root pointer is set to fs_root as was originally.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: use local fs_info in check_chunks_and_extents
David Sterba [Wed, 30 Aug 2017 17:41:15 +0000 (19:41 +0200)]
btrfs-progs: check: use local fs_info in check_chunks_and_extents

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: pass fs_info to check_chunks_and_extents
David Sterba [Wed, 30 Aug 2017 17:38:02 +0000 (19:38 +0200)]
btrfs-progs: check: pass fs_info to check_chunks_and_extents

The root pointer is set to fs_root as was originally.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: verify critical root pointers before use
David Sterba [Wed, 30 Aug 2017 17:25:11 +0000 (19:25 +0200)]
btrfs-progs: check: verify critical root pointers before use

The pointers to critical roots must be valid before we start using them,
eg. as the space clearing code.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: move more of space cache clearing to a helper
David Sterba [Wed, 30 Aug 2017 16:57:03 +0000 (18:57 +0200)]
btrfs-progs: check: move more of space cache clearing to a helper

Move the code out of main.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: remove old mixed backref code
David Sterba [Wed, 30 Aug 2017 14:56:51 +0000 (16:56 +0200)]
btrfs-progs: check: remove old mixed backref code

A code added in 2009 (95d3f20b51e9b) for a very short-lived change in
the format is no concern to us nowadays.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: add more sanitizer message patterns to log scanner
David Sterba [Wed, 30 Aug 2017 14:56:23 +0000 (16:56 +0200)]
btrfs-progs: tests: add more sanitizer message patterns to log scanner

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: convert: don't print message when fs is not recognized, ext2
David Sterba [Tue, 29 Aug 2017 17:42:10 +0000 (19:42 +0200)]
btrfs-progs: convert: don't print message when fs is not recognized, ext2

Bad magic error means it's not the desired filesystem so the error
message is just noise.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: limit size of log dump from conver tests
David Sterba [Tue, 29 Aug 2017 16:41:30 +0000 (18:41 +0200)]
btrfs-progs: tests: limit size of log dump from conver tests

The convert tests generate lots of log material, travis CI has limit 4MB
so we don't see anything useful when a late test fails.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: handle failed search in find_search_start better
David Sterba [Tue, 29 Aug 2017 16:19:03 +0000 (18:19 +0200)]
btrfs-progs: handle failed search in find_search_start better

The warning can pop up frequently on a fuzzed image, the message seems
to be enough. Add a more fitting error code too.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: add more error handling to btrfs_free_block_group
David Sterba [Tue, 29 Aug 2017 15:53:39 +0000 (17:53 +0200)]
btrfs-progs: add more error handling to btrfs_free_block_group

As btrfs_update_block_group fails when the block group is not found in
cache, we can exit btrfs_free_block_group, not much to rollback. The
caller will also exit in turn.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: fix TEST_LOG=dump in convert tests
David Sterba [Tue, 29 Aug 2017 15:42:28 +0000 (17:42 +0200)]
btrfs-progs: tests: fix TEST_LOG=dump in convert tests

Test failure in convert tests with log dump does not happen because
_fail is called before that and exits. Other test types are ok.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: handle transaction start failure in close_ctree
David Sterba [Tue, 29 Aug 2017 15:22:39 +0000 (17:22 +0200)]
btrfs-progs: handle transaction start failure in close_ctree

Closing the fs will try to commit a pending transaction, but may fail to
do so if the filesystem state is not well defined. This will eg.  fail
for some fuzz tests. The data structures are freed but no furhter
attempt to commit is made.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: update README
David Sterba [Tue, 29 Aug 2017 15:07:15 +0000 (17:07 +0200)]
btrfs-progs: tests: update README

Wording, runtime dependencies.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: install missing CI packages
David Sterba [Tue, 29 Aug 2017 14:36:15 +0000 (16:36 +0200)]
btrfs-progs: tests: install missing CI packages

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: add more checks for tools used in convert tests
David Sterba [Tue, 29 Aug 2017 14:35:01 +0000 (16:35 +0200)]
btrfs-progs: tests: add more checks for tools used in convert tests

Tools that may not be present in common installations should be checked
in the tests.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: build preparation phases out of script section
David Sterba [Tue, 29 Aug 2017 14:16:44 +0000 (16:16 +0200)]
btrfs-progs: tests: build preparation phases out of script section

The logs of before_install are folded and we don't need to see the
details, unlike the test logs.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: don't start or commit after transaction abort
David Sterba [Tue, 29 Aug 2017 12:55:37 +0000 (14:55 +0200)]
btrfs-progs: don't start or commit after transaction abort

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: start framework for transaction abort
David Sterba [Mon, 28 Aug 2017 15:44:41 +0000 (17:44 +0200)]
btrfs-progs: start framework for transaction abort

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: store pointer to fs_info in transaction handle
David Sterba [Mon, 28 Aug 2017 15:43:10 +0000 (17:43 +0200)]
btrfs-progs: store pointer to fs_info in transaction handle

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: move transaction code out of disk-io
David Sterba [Mon, 28 Aug 2017 15:39:26 +0000 (17:39 +0200)]
btrfs-progs: move transaction code out of disk-io

Temporarily export the low-level helpers.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: move transaction implementation out of header
David Sterba [Mon, 28 Aug 2017 15:27:08 +0000 (17:27 +0200)]
btrfs-progs: move transaction implementation out of header

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: use kzalloc to allocate transaction handle
David Sterba [Mon, 28 Aug 2017 15:21:05 +0000 (17:21 +0200)]
btrfs-progs: use kzalloc to allocate transaction handle

Use the kernel-style allocation helpers and remove redundant zeroing.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: switch fs_info::system_allocs to bit
David Sterba [Mon, 28 Aug 2017 15:16:51 +0000 (17:16 +0200)]
btrfs-progs: switch fs_info::system_allocs to bit

The value really is 0 or 1, the test in btrfs_reserve_extent can be
simplified.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: return errors from btrfs_start_transaction
David Sterba [Mon, 28 Aug 2017 14:54:12 +0000 (16:54 +0200)]
btrfs-progs: return errors from btrfs_start_transaction

Now that all callers will catch errors, we can return them from the
function.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: add crude error handling when transaction start fails
David Sterba [Mon, 28 Aug 2017 14:48:16 +0000 (16:48 +0200)]
btrfs-progs: add crude error handling when transaction start fails

Currently transaction bugs out insided btrfs_start_transaction in case
of error, we want to lift the error handling to the callers. This patch
adds the BUG_ON anywhere it's been missing so far. This is not the best
way of course. Transforming BUG_ON to a proper error handling highly
depends on the caller and should be dealt with case by case.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: pull and build reiserfs lib inside CI
David Sterba [Mon, 28 Aug 2017 12:28:53 +0000 (14:28 +0200)]
btrfs-progs: tests: pull and build reiserfs lib inside CI

The functionality needed to support reiserfs in convert has been added
recently and is not yet available in the CI images, so we'll download
and build it there.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: add build checks for sanitization features
David Sterba [Mon, 28 Aug 2017 11:54:02 +0000 (13:54 +0200)]
btrfs-progs: tests: add build checks for sanitization features

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: build: fix asan and ubsan build
David Sterba [Mon, 28 Aug 2017 11:38:32 +0000 (13:38 +0200)]
btrfs-progs: build: fix asan and ubsan build

Probably with a new gcc (7.1.1) I started to see asan/ubsan link failures.
Fixed by explicitly linking the libraries.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: build: fix building libbtrfs.so with tsan
David Sterba [Mon, 28 Aug 2017 11:33:19 +0000 (13:33 +0200)]
btrfs-progs: build: fix building libbtrfs.so with tsan

Except libbtrfs.so object, all other tools compile fine. The error is:

ld: send-stream.o: relocation R_X86_64_PC32 against symbol
`stderr@@GLIBC_2.2.5' can not be used when making a shared object;
recompile with -fPIC

Compiling with -fPIC fixes the problem.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: build: properly pass the LDFLAGS during tsan build
David Sterba [Fri, 25 Aug 2017 18:09:34 +0000 (20:09 +0200)]
btrfs-progs: build: properly pass the LDFLAGS during tsan build

The variable LD_FLAGS does not exist and the flags are not used, we need
to use the newly added internal debugging linker flags.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: build: add internal LDFLAGS for the D= features
David Sterba [Fri, 25 Aug 2017 18:04:48 +0000 (20:04 +0200)]
btrfs-progs: build: add internal LDFLAGS for the D= features

We might want to pass additional linker flags for various sanitization
features.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: add more fuzzed images from bugzilla
David Sterba [Fri, 25 Aug 2017 17:21:01 +0000 (19:21 +0200)]
btrfs-progs: tests: add more fuzzed images from bugzilla

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: handle blockgroup errors when opening filesystem
David Sterba [Fri, 25 Aug 2017 17:08:37 +0000 (19:08 +0200)]
btrfs-progs: handle blockgroup errors when opening filesystem

The call to btrfs_read_block_groups could loop if the metadata are
damaged (reported eg. for an unaligned block), due to lack of error
handling. We have to check for restored images or currently created
filesystems, that do not contain the blockgroups.

Can be reproduced by fuzzed image bko-155551.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=155551
Reported-by: Lukas Lueg <lukas.lueg@gmail.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: drop blocksize argument from readahead_tree_block
David Sterba [Fri, 25 Aug 2017 16:07:15 +0000 (18:07 +0200)]
btrfs-progs: drop blocksize argument from readahead_tree_block

Tree blocks are always nodesize. As readahead is only an optimization,
exact size is not required and is only advisory.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: drop blocksize argument from btrfs_find_create_tree_block
David Sterba [Fri, 25 Aug 2017 15:44:22 +0000 (17:44 +0200)]
btrfs-progs: drop blocksize argument from btrfs_find_create_tree_block

Metadata blocks are always nodesize. When reading the
superblock::sys_array, the actual size of data is fixed to 4k and
smaller than nodesize, but otherwise everything works as before.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: drop redundant check of blocksize in read_tree_block
David Sterba [Fri, 25 Aug 2017 15:30:34 +0000 (17:30 +0200)]
btrfs-progs: drop redundant check of blocksize in read_tree_block

The tree blocks are supposed to be always of nodesize. Before the
parameter has been dropped, it was unlikely but possible to pass a
misaligned value.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: cleanup use of level_size
David Sterba [Fri, 25 Aug 2017 15:21:03 +0000 (17:21 +0200)]
btrfs-progs: check: cleanup use of level_size

Nodesize is same for all levels, besides it's been only set and not
used, in root_item_record.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: drop blocksize from read_tree_block
David Sterba [Fri, 25 Aug 2017 14:54:16 +0000 (16:54 +0200)]
btrfs-progs: drop blocksize from read_tree_block

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: drop local blocksize variables if they're nodesize
David Sterba [Fri, 25 Aug 2017 14:20:16 +0000 (16:20 +0200)]
btrfs-progs: drop local blocksize variables if they're nodesize

Prep work so we can drop the blocksize argument from several functions.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: build: fix PIE build
David Sterba [Fri, 25 Aug 2017 12:26:51 +0000 (14:26 +0200)]
btrfs-progs: build: fix PIE build

Patch from Marcus Meissner <meissner@suse.com>. The CFLAGS are passed to
the linker and mix up the compilation and linker flags for PIE support.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: convert: move reiserfs struct definitions to header
David Sterba [Thu, 24 Aug 2017 13:48:48 +0000 (15:48 +0200)]
btrfs-progs: convert: move reiserfs struct definitions to header

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: convert: move and rename dev_t helpers to common file
David Sterba [Tue, 22 Aug 2017 18:43:49 +0000 (20:43 +0200)]
btrfs-progs: convert: move and rename dev_t helpers to common file

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: convert: move acl helper to common source file
David Sterba [Tue, 22 Aug 2017 18:34:36 +0000 (20:34 +0200)]
btrfs-progs: convert: move acl helper to common source file

There were 2 copies of ext2_acl_count and acl_ea_size.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: convert: move duplicated acl code to common header
David Sterba [Tue, 22 Aug 2017 18:30:43 +0000 (20:30 +0200)]
btrfs-progs: convert: move duplicated acl code to common header

The ACL types and macros are same for both source filesystem.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: docs: update btrfs-convert regarding reiserfs
David Sterba [Tue, 22 Aug 2017 17:18:48 +0000 (19:18 +0200)]
btrfs-progs: docs: update btrfs-convert regarding reiserfs

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: add support for converting reiserfs
Jeff Mahoney [Tue, 22 Aug 2017 16:32:57 +0000 (18:32 +0200)]
btrfs-progs: tests: add support for converting reiserfs

Many of the test cases for convert apply regardless of what the source
file system is and using ext4 is sufficient.  I've included several
test cases that are reiserfs-specific.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
[ patch split from the previous one, minor cleanups in common.convert ]
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: convert: add support for converting reiserfs
Jeff Mahoney [Tue, 22 Aug 2017 16:30:43 +0000 (18:30 +0200)]
btrfs-progs: convert: add support for converting reiserfs

This patch adds support to convert reiserfs file systems in-place to btrfs.

It will convert extended attribute files to btrfs extended attributes,
translate ACLs, coalesce tails that consist of multiple items into one item,
and convert tails that are too big into indirect files.

This requires that libreiserfscore 3.6.27 be available.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: Use named constants for common sizes
Nikolay Borisov [Thu, 27 Jul 2017 08:17:00 +0000 (11:17 +0300)]
btrfs-progs: Use named constants for common sizes

There multiple places where we use well-known sizes - 1,8,16,32 megabytes. We
also have them defined as constants in the sizes.h header. So let's use them.
No functional changes.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: mkfs: Replace number with enum
Gu Jinxiang [Wed, 28 Jun 2017 09:59:24 +0000 (17:59 +0800)]
btrfs-progs: mkfs: Replace number with enum

For code maintainability and scalability,
replace hardcoded constant with a meaningful enum.

Signed-off-by: Gu Jinxiang <gujx@cn.fujitsu.com>
[ add MKFS_ prefix ]
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: fix _is_file_or_command detection
David Sterba [Fri, 25 Aug 2017 13:44:24 +0000 (15:44 +0200)]
btrfs-progs: tests: fix _is_file_or_command detection

type -p returns an empty string for nonexistent commands, but the -f
test on an empty string does not behave the same on all shells. To be
safe, use the quoted value.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agoBtrfs progs v4.12.1
David Sterba [Fri, 25 Aug 2017 13:11:11 +0000 (15:11 +0200)]
Btrfs progs v4.12.1

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: update CHANGES for v4.12.1
David Sterba [Fri, 31 Mar 2017 12:20:21 +0000 (14:20 +0200)]
btrfs-progs: update CHANGES for v4.12.1

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: docs: add figure 1 to btrfs quota doc
Misono, Tomohiro [Thu, 24 Aug 2017 17:45:55 +0000 (19:45 +0200)]
btrfs-progs: docs: add figure 1 to btrfs quota doc

The document of btrfs quota is missing figure 1.

I notice the body is copy of http://sensille.com/qgroups.pdf (which is
linked from https://btrfs.wiki.kernel.org/index.php/Quota_support), and
insert the figure.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: docs: add missing short option for qroup-report
Misono, Tomohiro [Thu, 24 Aug 2017 06:06:41 +0000 (15:06 +0900)]
btrfs-progs: docs: add missing short option for qroup-report

Usage info of "btrfs check" shows "-Q|--qgroup-report" (and first patch
enables -Q), but the document only shows "--qgroup-report".

Therefore add -Q to the doc.

Signed-off-by: Tomohiro Misono <misono.tomohiro@jp.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: dump-super: add missing long option bytenr to getopt
Misono, Tomohiro [Thu, 24 Aug 2017 05:25:29 +0000 (14:25 +0900)]
btrfs-progs: dump-super: add missing long option bytenr to getopt

This fixes "btrfs inspect-internal dump-super --bytenr"

Signed-off-by: Tomohiro Misono <misono.tomohiro@jp.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: recieve: add missing short option E to getopt
Misono, Tomohiro [Thu, 24 Aug 2017 05:24:03 +0000 (14:24 +0900)]
btrfs-progs: recieve: add missing short option E to getopt

This fixes "btrfs receive -E".

Signed-off-by: Tomohiro Misono <misono.tomohiro@jp.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: add missing short options E and Q to getopt
Misono, Tomohiro [Thu, 24 Aug 2017 05:18:25 +0000 (14:18 +0900)]
btrfs-progs: check: add missing short options E and Q to getopt

I found some btrfs commands options are not working because of
inappropriate getopt_long() setting.

This fixes "btrfs check -Q/-E"

Signed-off-by: Tomohiro Misono <misono.tomohiro@jp.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: move command definitions to commands.h
David Sterba [Thu, 24 Aug 2017 16:32:18 +0000 (18:32 +0200)]
btrfs-progs: move command definitions to commands.h

There are some trivial helpers, we can group the command declarations in
one place.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: catch bad usage of run_mustfail
David Sterba [Wed, 23 Aug 2017 17:26:42 +0000 (19:26 +0200)]
btrfs-progs: tests: catch bad usage of run_mustfail

This function has an extra argument and can get forgotten, add a sanity
check so the bad usage can be caught during development.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: scan results for commands not found
David Sterba [Wed, 23 Aug 2017 17:19:47 +0000 (19:19 +0200)]
btrfs-progs: tests: scan results for commands not found

In case of typos or messed up command execution, we'd like to be able to
catch that.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: fsck-test: case for corrupted dir item name
Su Yue [Fri, 14 Jul 2017 07:47:46 +0000 (15:47 +0800)]
btrfs-progs: fsck-test: case for corrupted dir item name

In this test case, all name in dir_item, dir_index, inode_ref
are corrupted to another one.
btrfs check should report errors about the corrupted dir_item but
btrfs can't repair the case now.

Signed-off-by: Su Yue <suy.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: verify dir item name and hash in lowmem mode
Su Yue [Fri, 14 Jul 2017 07:47:45 +0000 (15:47 +0800)]
btrfs-progs: check: verify dir item name and hash in lowmem mode

Although lowmem mode can detect name and hash mismatch in dir_item,
it's done by checking inode_ref to expose such problem.

This patch will enhance dir_item check, by also comparing name and
hash when checking dir_items.

Reported-by: Filippe LeMarchand <gasinvein@gmail.com>
Signed-off-by: Su Yue <suy.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: verify name in dir_item for original mode
Su Yue [Fri, 14 Jul 2017 07:47:44 +0000 (15:47 +0800)]
btrfs-progs: check: verify name in dir_item for original mode

In original mode, we don't check if the name in dir_item matches the
hash in key.offset.

In the following case, original mode will report nothing wrong while
lowmem mode will detect the name and hash mismatch.

------
item 72 key (79177 DIR_ITEM 54846528) itemoff 12380 itemsize 88
        location key (4222342 INODE_ITEM 0) type FILE
        transid 170929 data_len 0 name_len 14
        name: deprecated.sxt
        location key (13590433 INODE_ITEM 0) type FILE
        transid 796448 data_len 0 name_len 14
        name: deprecated.txt
------
In above case, hash of "deprecated.txt" matches with 54846528,
while hash of "deprecated.sxt" should be 2008317993.

Reported-by: Filippe LeMarchand <gasinvein@gmail.com>
Signed-off-by: Su Yue <suy.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: fix typo in convert-tests/008-readonly-image
Jeff Mahoney [Tue, 22 Aug 2017 17:44:44 +0000 (19:44 +0200)]
btrfs-progs: tests: fix typo in convert-tests/008-readonly-image

The dd in convert-tests/008-readonly-image is expected to fail, so
there being a typo in the file name has gone unnoticed.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: convert: use search_cache_extent in migrate_one_reserved_range
Jeff Mahoney [Thu, 27 Jul 2017 15:47:21 +0000 (11:47 -0400)]
btrfs-progs: convert: use search_cache_extent in migrate_one_reserved_range

When we are looking for extents in migrate_one_reserved_range, it's likely
that there will be multiple extents that fall into the 0-1MB range.

If lookup_cache_extent is called with a range that covers multiple cache
entries, it will return the first entry it encounters while searching
from the top of the tree that happens to fall in that range.  That
means that we can end up skipping regions within that range, resulting
in a file system image that can't be rolled back since it wasn't
all migrated properly.

This is reproducible using convert-tests/008-readonly-image.  There was
a range from 0-160kB, but the only entry that was returned began at
~ 280kB.

The fix is to use search_cache_extent to iterate through multiple regions
within that range.

Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Reviewed-by: Qu Wenruo <quwenruo.btrfs@gmx.com>
Signed-off-by: David Sterba <dsterba@suse.com>