platform/upstream/btrfs-progs.git
6 years agobtrfs-progs: Fix restoring image from multi devices fs into single device
Filipe Manana [Wed, 17 May 2017 01:41:21 +0000 (02:41 +0100)]
btrfs-progs: Fix restoring image from multi devices fs into single device

We correctly build an image from a multiple devices filesystem but when
restoring the image into a single device we were missing updating the
number of devices in the superblock to the value 1 (we already took care
of setting the number of stripes to 1 for each chunk item and setting
the device id for each chunk item to match the device id from the super
block).

This missing update of the number of devices makes it impossible to mount
the restored filesystem on recent kernels, more specifically since the
linux kernel commit 99e3ecfcb9f4ca35192d20a5bea158b81f600062
("Btrfs: add more validation checks for superblock"), that produce the
following message in the dmesg/syslog:

[21097.542047] BTRFS error (device sdi): super_num_devices 2 mismatch with num_devices 1 found here
[21097.543972] BTRFS error (device sdi): failed to read chunk tree: -22
[21097.720360] BTRFS error (device sdi): open_ctree failed

So fix this by updating the number of devices to 1 in the superblock.

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>
6 years agobtrfs-progs: tests: remove variable quotation from convert-tests
Tsutomu Itoh [Tue, 16 May 2017 03:01:53 +0000 (12:01 +0900)]
btrfs-progs: tests: remove variable quotation from convert-tests

In btrfs-progs-v4.11-rc1, the following convert-tests failed.

    [TEST/conv]   008-readonly-image
    [TEST/conv]     readonly image test, btrfs defaults
failed: mke2fs -t ext4 -b 4096 -F /Build/btrfs-progs-v4.11-rc1/tests/test.img
test failed for case 008-readonly-image
Makefile:271: recipe for target 'test-convert' failed
make: *** [test-convert] Error 1
    [TEST/conv]   009-common-inode-flags
    [TEST/conv]     common inode flags test, btrfs defaults
failed: mke2fs -t ext4 -b 4096 -F /Build/btrfs-progs-v4.11-rc1/tests/test.img
test failed for case 009-common-inode-flags
Makefile:271: recipe for target 'test-convert' failed
make: *** [test-convert] Error 1

So, remove quotes from $default_mke2fs.

Signed-off-by: Tsutomu Itoh <t-itoh@jp.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agoBtrfs progs v4.11
David Sterba [Thu, 18 May 2017 14:16:03 +0000 (16:16 +0200)]
Btrfs progs v4.11

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: update CHANGES for v4.11
David Sterba [Fri, 31 Mar 2017 12:20:21 +0000 (14:20 +0200)]
btrfs-progs: update CHANGES for v4.11

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: fix misc/019-receive-clones-on-munted-subvol
David Sterba [Mon, 15 May 2017 14:34:31 +0000 (16:34 +0200)]
btrfs-progs: tests: fix misc/019-receive-clones-on-munted-subvol

Patches "btrfs-progs: tests: correctly receive clones to mounted subvol"
(8eaf63bc9a7b957c566be23df7c6701a9a5b22a2) and followup are missing last
unmount which leads to failure of misc/020.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: receiv: fail on first -ENODATA only
Christian Brauner [Sat, 29 Apr 2017 21:54:05 +0000 (23:54 +0200)]
btrfs-progs: receiv: fail on first -ENODATA only

Returning -ENODATA is only considered invalid on the first run of the
loop where we would detect entirely empty stream.

The enhanced test misc-tests/018-recv-end-of-stream now passes.

Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=195597
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: print-tree: Add leaf flags and backref revision output
Qu Wenruo [Mon, 8 May 2017 07:38:10 +0000 (15:38 +0800)]
btrfs-progs: print-tree: Add leaf flags and backref revision output

Btrfs header has a u64 member flags, whose lowest 56 bits are for header
flags like WRITTEN and RELOC.
And its highest 8 bits are for backref revision.

Manually checking btrfs_header_flags() will be a pain, so add such leaf
flags and backref revision output for print-tree.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: Avoid reading beyond item boundary for dir_item and dir_index
Qu Wenruo [Wed, 3 May 2017 08:42:40 +0000 (16:42 +0800)]
btrfs-progs: check: Avoid reading beyond item boundary for dir_item and dir_index

When reading out name from inode_ref, it's possible that corrupted
name_len can lead to read beyond boundary of item or even extent buffer.

This happens when checking fuzzed image /tmp/bko-161811.raw, for both
lowmem mode and original mode.

Below is the example from lowmem mode.

ERROR: root 5 INODE REF[256 256] doesn't have related DIR_INDEX[256 216172782113783808] namelen 255 filename bar filetype 0
ERROR: root 5 INODE REF[256 256] doesn't have related DIR_ITEM[256 1306590535] namelen 255 filename bar filetype 0
WARNING: root 5 INODE[256] mode 0 shouldn't have DIR_INDEX[256 1167283096]
WARNING: root 5 DIR_ITEM[256 1167283096] name too long
==13013== Invalid read of size 1
==13013==    at 0x4C31A38: memmove (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==13013==    by 0x431518: read_extent_buffer (extent_io.c:863)
==13013==    by 0x4752AB: check_dir_item (cmds-check.c:4627)
==13013==    by 0x475E5C: check_inode_item (cmds-check.c:4911)
==13013==    by 0x476200: check_fs_first_inode (cmds-check.c:5011)
==13013==    by 0x476276: check_fs_root_v2 (cmds-check.c:5044)
==13013==    by 0x4769FB: check_fs_roots_v2 (cmds-check.c:5242)
==13013==    by 0x488B5B: cmd_check (cmds-check.c:13033)
==13013==    by 0x40A8C5: main (btrfs.c:246)
==13013==  Address 0x5c95b80 is 0 bytes after a block of size 4,224 alloc'd
==13013==    at 0x4C2CF35: calloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==13013==    by 0x4307E0: __alloc_extent_buffer (extent_io.c:538)
==13013==    by 0x430C37: alloc_extent_buffer (extent_io.c:642)
==13013==    by 0x413DFE: btrfs_find_create_tree_block (disk-io.c:193)
==13013==    by 0x414370: read_tree_block_fs_info (disk-io.c:340)
==13013==    by 0x40B5D5: read_tree_block (disk-io.h:125)
==13013==    by 0x40CFD2: read_node_slot (ctree.c:652)
==13013==    by 0x40E5EB: btrfs_search_slot (ctree.c:1172)
==13013==    by 0x4761A8: check_fs_first_inode (cmds-check.c:5001)
==13013==    by 0x476276: check_fs_root_v2 (cmds-check.c:5044)
==13013==    by 0x4769FB: check_fs_roots_v2 (cmds-check.c:5242)
==13013==    by 0x488B5B: cmd_check (cmds-check.c:13033)

Fix it by double checking dir_item, name_len against item boundary
before trying to read out name from extent buffer, for both original
mode and lowmem mode.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: Avoid reading beyond item boundary for inode_ref
Qu Wenruo [Wed, 3 May 2017 08:42:39 +0000 (16:42 +0800)]
btrfs-progs: check: Avoid reading beyond item boundary for inode_ref

When reading out name from inode_ref, it's possible that corrupted
name_len can lead to read beyond boundary of item or even extent buffer.

This happens when checking fuzzed image /tmp/bko-161811.raw, for both
lowmem mode and original mode.

ERROR: root 5 INODE REF[256 256] doesn't have related DIR_INDEX[256 504403158265495680] namelen 0 filename  filetype 0
ERROR: root 5 INODE REF[256 256] doesn't have related DIR_ITEM[256 4294967294] namelen 0 filename  filetype 0
WARNING: root 5 INODE_REF[256 256] name too long
==13022== Invalid read of size 8
==13022==    at 0x4C319BE: memmove (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==13022==    by 0x431518: read_extent_buffer (extent_io.c:863)
==13022==    by 0x474730: check_inode_ref (cmds-check.c:4307)
==13022==    by 0x475D65: check_inode_item (cmds-check.c:4890)
==13022==    by 0x476200: check_fs_first_inode (cmds-check.c:5011)
==13022==    by 0x476276: check_fs_root_v2 (cmds-check.c:5044)
==13022==    by 0x4769FB: check_fs_roots_v2 (cmds-check.c:5242)
==13022==    by 0x488B5B: cmd_check (cmds-check.c:13033)
==13022==    by 0x40A8C5: main (btrfs.c:246)
==13022==  Address 0x5c96780 is 0 bytes after a block of size 4,224 alloc'd
==13022==    at 0x4C2CF35: calloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==13022==    by 0x4307E0: __alloc_extent_buffer (extent_io.c:538)
==13022==    by 0x430C37: alloc_extent_buffer (extent_io.c:642)
==13022==    by 0x413DFE: btrfs_find_create_tree_block (disk-io.c:193)
==13022==    by 0x414370: read_tree_block_fs_info (disk-io.c:340)
==13022==    by 0x40B5D5: read_tree_block (disk-io.h:125)
==13022==    by 0x40CFD2: read_node_slot (ctree.c:652)
==13022==    by 0x40E5EB: btrfs_search_slot (ctree.c:1172)
==13022==    by 0x4761A8: check_fs_first_inode (cmds-check.c:5001)
==13022==    by 0x476276: check_fs_root_v2 (cmds-check.c:5044)
==13022==    by 0x4769FB: check_fs_roots_v2 (cmds-check.c:5242)
==13022==    by 0x488B5B: cmd_check (cmds-check.c:13033)
=

Fix it by double checking inode_ref, name_len against item boundary
before trying to read out name from extent buffer, for both original
mode and lowmem mode.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: Fix heap use after free
Su Yue [Wed, 3 May 2017 01:50:14 +0000 (09:50 +0800)]
btrfs-progs: check: Fix heap use after free

fsck/004-no-dir-index makes valgrinds complaining about Invalid read.
==31890== Invalid read of size 1
==31890==    at 0x453D09: repair_inode_backrefs (cmds-check.c:2690)
==31890==    by 0x453D09: check_inode_recs (cmds-check.c:3330)
==31890==    by 0x453D09: check_fs_root (cmds-check.c:4012)
==31890==    by 0x45E788: check_fs_roots (cmds-check.c:4098)
==31890==    by 0x45E788: cmd_check (cmds-check.c:13031)
==31890==    by 0x40A88A: main (btrfs.c:246)
==31890==  Address 0x5cb7b90 is 16 bytes inside a block of size 50 free'd
==31890==    at 0x4C2C14B: free (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==31890==    by 0x453D08: repair_inode_backrefs (cmds-check.c:2684)
==31890==    by 0x453D08: check_inode_recs (cmds-check.c:3330)
==31890==    by 0x453D08: check_fs_root (cmds-check.c:4012)
==31890==    by 0x45E788: check_fs_roots (cmds-check.c:4098)
==31890==    by 0x45E788: cmd_check (cmds-check.c:13031)
==31890==    by 0x40A88A: main (btrfs.c:246)
==31890==  Block was alloc'd at
==31890==    at 0x4C2AF1F: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==31890==    by 0x45055C: get_inode_backref (cmds-check.c:1075)
==31890==    by 0x45055C: add_inode_backref (cmds-check.c:1097)
==31890==    by 0x45180C: process_dir_item (cmds-check.c:1525)
==31890==    by 0x45180C: process_one_leaf (cmds-check.c:1838)
==31890==    by 0x45180C: walk_down_tree (cmds-check.c:2134)
==31890==    by 0x45180C: check_fs_root (cmds-check.c:3957)
==31890==    by 0x45E788: check_fs_roots (cmds-check.c:4098)
==31890==    by 0x45E788: cmd_check (cmds-check.c:13031)
==31890==    by 0x40A88A: main (btrfs.c:246)
==31890==
==31890== Invalid read of size 8
==31890==    at 0x452D66: repair_inode_backrefs (cmds-check.c:2731)
==31890==    by 0x452D66: check_inode_recs (cmds-check.c:3330)
==31890==    by 0x452D66: check_fs_root (cmds-check.c:4012)
==31890==    by 0x45E788: check_fs_roots (cmds-check.c:4098)
==31890==    by 0x45E788: cmd_check (cmds-check.c:13031)
==31890==    by 0x40A88A: main (btrfs.c:246)
==31890==  Address 0x5cb7b90 is 16 bytes inside a block of size 50 free'd
==31890==    at 0x4C2C14B: free (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==31890==    by 0x453D08: repair_inode_backrefs (cmds-check.c:2684)
==31890==    by 0x453D08: check_inode_recs (cmds-check.c:3330)
==31890==    by 0x453D08: check_fs_root (cmds-check.c:4012)
==31890==    by 0x45E788: check_fs_roots (cmds-check.c:4098)
==31890==    by 0x45E788: cmd_check (cmds-check.c:13031)
==31890==    by 0x40A88A: main (btrfs.c:246)
==31890==  Block was alloc'd at
==31890==    at 0x4C2AF1F: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==31890==    by 0x45055C: get_inode_backref (cmds-check.c:1075)
==31890==    by 0x45055C: add_inode_backref (cmds-check.c:1097)
==31890==    by 0x45180C: process_dir_item (cmds-check.c:1525)
==31890==    by 0x45180C: process_one_leaf (cmds-check.c:1838)
==31890==    by 0x45180C: walk_down_tree (cmds-check.c:2134)
==31890==    by 0x45180C: check_fs_root (cmds-check.c:3957)
==31890==    by 0x45E788: check_fs_roots (cmds-check.c:4098)
==31890==    by 0x45E788: cmd_check (cmds-check.c:13031)
==31890==    by 0x40A88A: main (btrfs.c:246)
==31890==

While iterating over backrefs in repair_inode_backrefs, there are
several situations to repair one backref according
backref->found_dir_item and backref->found_dir_index.  Two of these
branches may free the backref, but next checks will still access the
freed memory.

Because these branches are independent, let repair_inode_backrefs skip
to handle next backref after free can fix it.

Signed-off-by: Su Yue <suy.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: add variable quotation to convert-tests
Lakshmipathi.G [Sun, 23 Apr 2017 19:00:00 +0000 (00:30 +0530)]
btrfs-progs: tests: add variable quotation to convert-tests

Signed-off-by: Lakshmipathi.G <Lakshmipathi.G@giis.co.in>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: make max_size consistent with nr
Christophe de Dinechin [Fri, 28 Apr 2017 09:51:21 +0000 (11:51 +0200)]
btrfs-progs: check: make max_size consistent with nr

Since we memset tmpl, max_size==0. This does not seem consistent with nr = 1.
In check_extent_refs, we will call:

  set_extent_dirty(root->fs_info->excluded_extents,
                   rec->start,
                   rec->start + rec->max_size - 1);

This ends up with BUG_ON(end < start) in insert_state.

Signed-off-by: Christophe de Dinechin <dinechin@redhat.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: prevent attempt to insert extent record with max_size==0
Christophe de Dinechin [Fri, 28 Apr 2017 09:50:23 +0000 (11:50 +0200)]
btrfs-progs: check: prevent attempt to insert extent record with max_size==0

When this happens, we will trip a BUG_ON(end < start) in insert_state
because in check_extent_refs, we use this max_size expecting it's not zero:

  set_extent_dirty(root->fs_info->excluded_extents,
                   rec->start,
                   rec->start + rec->max_size - 1);

See https://bugzilla.redhat.com/show_bug.cgi?id=1435567
for an example where this scenario occurs.

Signed-off-by: Christophe de Dinechin <dinechin@redhat.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: disambiguate between cases where add_tree_backref fails
Christophe de Dinechin [Fri, 28 Apr 2017 09:12:35 +0000 (11:12 +0200)]
btrfs-progs: check: disambiguate between cases where add_tree_backref fails

See https://bugzilla.redhat.com/show_bug.cgi?id=1435567 for an example
where the message occurs.

Signed-off-by: Christophe de Dinechin <dinechin@redhat.com>
[ un-indent strings overfowing 80 cols ]
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: Use more strict check to read out tree root
Qu Wenruo [Tue, 25 Apr 2017 08:40:16 +0000 (16:40 +0800)]
btrfs-progs: Use more strict check to read out tree root

Fuzzed image bko-156811-bad-parent-ref-qgroup-verify.raw causes qgroup
to report -ENOMEM.

But the fact is, such image is heavily damaged so there is no valid root
item for the extent tree.

Normal extent tree key in root tree should be (EXTENT_TREE ROOT_ITEM 0),
while in that fuzzed image, we got (EXTENT_TREE EXXTENT_DATA SOME_NUMBER).

It's btrfs_find_last_root() that only checks the objectid, not caring
about the key type leading to such problem.

Fix it by doing extra check on key type.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
[ edit changelog ]
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: Fix fuzz-test for bko-161821.raw.txt
Lu Fengqi [Tue, 2 May 2017 07:36:09 +0000 (15:36 +0800)]
btrfs-progs: tests: Fix fuzz-test for bko-161821.raw.txt

Fuzzed image bko-161821.raw causes btrfs check to get segmentation fault.

The function check_owner_ref attempts to access a non-exist quota tree
when dealing with extent_item [4198400 4096] in the corrupted filesystem.

The function btrfs_new_fs_info always allocates memory for
fs_info->quota_root regardless of whether quota_tree exists or not.
Additionally, the function btrfs_read_fs_root will directly return
fs_info->quota_root if location->objectid == BTRFS_QUOTA_TREE_OBJECTID.

This patch does the following things:
1. Do extra check and return ENOENT if quota tree does not exist in the
   function btrfs_read_fs_root.
2. Free useless fs_info->quota_root in the function btrfs_setup_all_roots
   to reduce confusion.
3. free_extent_buffer even if check_child_node failed in the function
   walk_down_tree.

Signed-off-by: Lu Fengqi <lufq.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: Fix memory leak in check_chunks_and_extents
Qu Wenruo [Tue, 2 May 2017 03:17:25 +0000 (11:17 +0800)]
btrfs-progs: check: Fix memory leak in check_chunks_and_extents

fsck/003-shift-offsets makes valgrinds complaining about memory leaks.
==5910==
==5910== HEAP SUMMARY:
==5910==     in use at exit: 1,112 bytes in 11 blocks
==5910==   total heap usage: 161 allocs, 150 frees, 164,800 bytes allocated
==5910==
==5910== 216 (72 direct, 144 indirect) bytes in 1 blocks are definitely lost in loss record 3 of 5
==5910==    at 0x4C2AF1F: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==5910==    by 0x4815A3: add_root_item_to_list (cmds-check.c:9683)
==5910==    by 0x481CE2: check_chunks_and_extents (cmds-check.c:9886)
==5910==    by 0x48888B: cmd_check (cmds-check.c:12977)
==5910==    by 0x40A8C5: main (btrfs.c:246)
==5910==

The check_chunks_and_extents() memory leaks are caused by not freeing
added root items of normal_trees and dropping_trees.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: fssum: fix warning, include correct header for major()
David Sterba [Tue, 2 May 2017 14:23:54 +0000 (16:23 +0200)]
btrfs-progs: fssum: fix warning, include correct header for major()

tests/fssum.c:599:13: warning: In the GNU C Library, "major" is defined
 by <sys/sysmacros.h>. For historical compatibility, it is
 currently defined by <sys/types.h> as well, but we plan to
 remove this soon. To use "major", include <sys/sysmacros.h>
 directly. If you did not intend to use a system-defined macro
 "major", you should undefine it after including <sys/types.h>.
    sum_add_u64(&cs, major(st.st_rdev));

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: fssum, fix memory leak
Lu Fengqi [Wed, 26 Apr 2017 01:22:41 +0000 (09:22 +0800)]
btrfs-progs: tests: fssum, fix memory leak

Free the alloced memory and close dir before exit.

Signed-off-by: Lu Fengqi <lufq.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: Fix memory leak when 0 sized block group item is found
Qu Wenruo [Tue, 25 Apr 2017 08:01:17 +0000 (16:01 +0800)]
btrfs-progs: Fix memory leak when 0 sized block group item is found

When a 0 sized block group item is found, set_extent_bits() will not
really set any bits.
While set_state_private() still inserts allocated block group cache into
block group extent_io_tree.

So at close_ctree() time, we won't free the private block group cache
stored since we can't find any bit set for the 0 sized block group.

To fix it, at btrfs_read_block_groups() we skip any 0 sized block group,
so such leak won't happen.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: add variable quotation to fsck-tests
Lakshmipathi.G [Mon, 24 Apr 2017 13:29:40 +0000 (18:59 +0530)]
btrfs-progs: tests: add variable quotation to fsck-tests

Signed-off-by: Lakshmipathi.G <Lakshmipathi.G@giis.co.in>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: docs: update formatting and wording for btrfs(5)
David Sterba [Thu, 20 Apr 2017 16:50:02 +0000 (18:50 +0200)]
btrfs-progs: docs: update formatting and wording for btrfs(5)

- emphasize features in the list
- minor whitespace fixes

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: print-tree: add validation to print_chunk
Lu Fengqi [Thu, 20 Apr 2017 08:07:57 +0000 (16:07 +0800)]
btrfs-progs: print-tree: add validation to print_chunk

In print_chunk, validate the value of uuid_offset when read the dev_uuid of
stripe.

Was triggered by misc-test/015-dump-super-garbage running indefinetelly.

Issue: #37
Signed-off-by: Lu Fengqi <lufq.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: dump-super: check array_size in print_sys_chunk_array
Lu Fengqi [Thu, 20 Apr 2017 08:07:56 +0000 (16:07 +0800)]
btrfs-progs: dump-super: check array_size in print_sys_chunk_array

Without validation of array_size, the dump-super may lead to a bad
memory access.

Signed-off-by: Lu Fengqi <lufq.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: send: always print a space after path in dump
Evan Danaher [Tue, 11 Apr 2017 16:33:40 +0000 (12:33 -0400)]
btrfs-progs: send: always print a space after path in dump

I was shocked to discover that 'btrfs receive --dump' doesn't print a
space after long filenames, so it runs together into the metadata; for
example:

truncate        ./20-00-03/this-name-is-32-characters-longsize=0

This is a trivial patch to add a single space unconditionally, so the
result is the following:

truncate        ./20-00-03/this-name-is-32-characters-long size=0

I suppose this is technically a breaking change, but it seems unlikely
to me that anyone would depend on the existing behavior given how
unfriendly it is.

Signed-off-by: Evan Danaher <github@edanaher.net>
Reviewed-by: Noah Massey <noah.massey@gmail.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: misc-tests: Superblock corruption and recovery using backup
Lakshmipathi.G [Sat, 15 Apr 2017 10:13:08 +0000 (15:43 +0530)]
btrfs-progs: misc-tests: Superblock corruption and recovery using backup

Signed-off-by: Lakshmipathi.G <Lakshmipathi.G@giis.co.in>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: docs: Fix newlines for man btrfstune
Hans van Kranenburg [Sun, 16 Apr 2017 17:20:02 +0000 (19:20 +0200)]
btrfs-progs: docs: Fix newlines for man btrfstune

A bunch of newlines were missing, which resulted in only -S and -r to
show as option after xmlto is used to convert the documentation to a man
page.

The rest of the options would end up being appended to the explanation
of -r.

Signed-off-by: Hans van Kranenburg <hans.van.kranenburg@mendix.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: balance: minor wording adjustment for full balance warning
David Sterba [Thu, 13 Apr 2017 21:56:55 +0000 (23:56 +0200)]
btrfs-progs: balance: minor wording adjustment for full balance warning

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: docs: Fix missing newline in man 5 btrfs
Hans van Kranenburg [Sat, 8 Apr 2017 22:37:47 +0000 (00:37 +0200)]
btrfs-progs: docs: Fix missing newline in man 5 btrfs

The text compress_lzo:: would show up directly after 'bigger than the
page size' on the same line.

Signed-off-by: Hans van Kranenburg <hans.van.kranenburg@mendix.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: misc/018-receive use receive -e to terminate on end marker
Christian Brauner [Mon, 3 Apr 2017 20:21:08 +0000 (22:21 +0200)]
btrfs-progs: tests: misc/018-receive use receive -e to terminate on end marker

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
[ enhance tests to take extra options and use for -e ]
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: fix btrfs send & receive with -e flag
Christian Brauner [Mon, 3 Apr 2017 20:21:07 +0000 (22:21 +0200)]
btrfs-progs: fix btrfs send & receive with -e flag

The old check here tried to ensure that empty streams are not considered valid.
The old check however, will always fail when only one run through the while(1)
loop is needed and honor_end_cmd is set. So this:

btrfs send /some/subvol | btrfs receive -e /some/

will consistently fail because -e causes honor_cmd_to be set and
btrfs_read_and_process_send_stream() to correctly return 1. So the command will
be successful but btrfs receive will error out because the send - receive
concluded in one run through the while(1) loop.

If we want to exclude empty streams we need a way to tell the difference between
btrfs_read_and_process_send_stream() returning 1 because read_buf() did not
detect any data and read_and_process_cmd() returning 1 because honor_end_cmd was
set. Without introducing too many changes the best way to me seems to have
btrfs_read_and_process_send_stream() return -ENODATA in the first case. The rest
stays the same. We can then check for -ENODATA in do_receive() and report a
proper error in this case. This should also be backwards compatible to previous
versions of btrfs receive. They will fail on empty streams because a negative
value is returned. The only thing that they will lack is a nice error message.

Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: docs: btrfs-replace, fix typo
David Sterba [Fri, 7 Apr 2017 16:21:07 +0000 (12:21 -0400)]
btrfs-progs: docs: btrfs-replace, fix typo

Github pull request #39, from gabo77.

Author: 3psus <gabo77@me.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: README: sort bug reports means by preference
David Sterba [Mon, 3 Apr 2017 17:47:10 +0000 (19:47 +0200)]
btrfs-progs: README: sort bug reports means by preference

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: rework option parser to use getopt for global options
David Sterba [Mon, 21 Nov 2016 12:40:43 +0000 (13:40 +0100)]
btrfs-progs: rework option parser to use getopt for global options

Preparatory work to support more global options. The current parser
abuses the subcommand table to understand help and version when
specified as options (--). These are now special case when processing
the global options.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: docs: mount options, enhance ssd/nossd
David Sterba [Fri, 31 Mar 2017 17:40:05 +0000 (19:40 +0200)]
btrfs-progs: docs: mount options, enhance ssd/nossd

Reported-by: Hans van Kranenburg <hans.van.kranenburg@mendix.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agoBtrfs progs v4.10.2
David Sterba [Fri, 31 Mar 2017 15:29:12 +0000 (17:29 +0200)]
Btrfs progs v4.10.2

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: update CHANGES for v4.10.2
David Sterba [Fri, 31 Mar 2017 12:20:21 +0000 (14:20 +0200)]
btrfs-progs: update CHANGES for v4.10.2

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: fix missing __error symbol in libbtrfs.so.0
Sergei Trofimovich [Mon, 27 Mar 2017 21:07:20 +0000 (22:07 +0100)]
btrfs-progs: fix missing __error symbol in libbtrfs.so.0

The easiest way to reproduce the error is to try to build
btrfs-progs with
    $ make LDFLAGS=-Wl,--no-undefined

    btrfs-list.o: In function `lookup_ino_path':
    btrfs-list.c:(.text+0x7d2): undefined reference to `__error'

Noticed by Denis Descheneaux when snapper tool
stopped working after upgrade to btrfs-progs-4.10.

As soname didn't change in 4.9 -> 4.10 release
I assume it's just an object file omission
in library depends and not the API/ABI change
of the library error printing.

Cc: linux-btrfs@vger.kernel.org
Cc: Mike Gilbert <floppym@gentoo.org>
Reported-by: Denis Descheneaux
Bug: https://bugs.gentoo.org/show_bug.cgi?id=613890

Signed-off-by: Sergei Trofimovich <slyfox@gentoo.org>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: Add SHARED_DATA_REF test image for check lowmem mode
Qu Wenruo [Fri, 17 Mar 2017 02:06:34 +0000 (10:06 +0800)]
btrfs-progs: tests: Add SHARED_DATA_REF test image for check lowmem mode

Introduce a new image, which contains external SHARED_DATA_REF items to
trigger a lowmem mode false alert.

The image only contains external SHARED_DATA_REF and no inlined data
backref.

Before the image, we only have inlined shared data ref, which is not
enough to trigger lowmem mode false alert.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: lowmem, fix false alert about backref lost for SHARED_DATA_REF
Qu Wenruo [Fri, 17 Mar 2017 02:06:33 +0000 (10:06 +0800)]
btrfs-progs: check: lowmem, fix false alert about backref lost for SHARED_DATA_REF

In check_extent_data_item(), after checking extent item of one data
extent, we search inlined data backref, then EXTENT_DATA_REF_KEY.

But we didn't search SHARED_DATA_REF, so if the backref is
SHARED_DATA_REF, then we will raise a false alert about backref lost.

Fix by also checking SHARED_DATA_REF_KEY in check_extent_data_item().

Reported-by: Chris Murphy <chris@colorremedies.com>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: convert: Add missing return for HOLE mode when checking convert image
Qu Wenruo [Tue, 21 Mar 2017 01:00:55 +0000 (09:00 +0800)]
btrfs-progs: convert: Add missing return for HOLE mode when checking convert image

In check_convert_image(), for normal HOLE case, if the file extents are
smaller than image size, we set ret to -EINVAL and print error message.

But forget to return.

This patch adds the missing return to fix it.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: add prefix to message helpers
David Sterba [Mon, 27 Mar 2017 12:14:20 +0000 (14:14 +0200)]
btrfs-progs: add prefix to message helpers

The implementation of message helpers use very generic names so we
should at least use a prefix, as they're going to be usied from within
the library. The build fix will follow.

Reported-by: Mike Gilbert <floppym@gentoo.org>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: Cleanup kernel-shared dir when execute make clean
Qu Wenruo [Mon, 27 Mar 2017 00:52:09 +0000 (08:52 +0800)]
btrfs-progs: Cleanup kernel-shared dir when execute make clean

Reported-by: Chris <coderight@gmail.com>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: library-test: add all exported headers
David Sterba [Mon, 27 Mar 2017 11:55:01 +0000 (13:55 +0200)]
btrfs-progs: library-test: add all exported headers

Explicitly include all headers exported to library, as a light build
check that the files are really present.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agoBtrfs progs v4.10.1
David Sterba [Fri, 17 Mar 2017 13:11:22 +0000 (14:11 +0100)]
Btrfs progs v4.10.1

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: update CHANGES for v4.10.1
David Sterba [Thu, 16 Mar 2017 16:11:20 +0000 (17:11 +0100)]
btrfs-progs: update CHANGES for v4.10.1

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: convert: Make btrfs_reserved_ranges const
Qu Wenruo [Thu, 16 Mar 2017 03:18:31 +0000 (11:18 +0800)]
btrfs-progs: convert: Make btrfs_reserved_ranges const

Since btrfs_reserved_ranges array is just used to store btrfs reserved
ranges, no one will nor should modify them at run time, make them static
and const will be better.

This also eliminates the use of immediate number 3.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
[ definition stays in source-fs.c ]
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: kerncompat: Fix re-definition of __bitwise
Qu Wenruo [Thu, 16 Mar 2017 03:18:30 +0000 (11:18 +0800)]
btrfs-progs: kerncompat: Fix re-definition of __bitwise

In latest linux api headers, __bitwise is already defined in
/usr/include/linux/types.h.

So kerncompat.h will re-define __bitwise, and cause gcc warning.

Fix it by checking if __bitwise is already define.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: dump-tree: Also output log root tree
Qu Wenruo [Tue, 14 Mar 2017 05:35:18 +0000 (13:35 +0800)]
btrfs-progs: dump-tree: Also output log root tree

In btrfs-dump-tree, we output any existing log tree, however we don't
output the log root tree, which records all root items for log trees.

This makes it confusing for any one who want to know where the log tree
comes from.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: dump-tree: Fix duplicated output when using -t option
Qu Wenruo [Tue, 14 Mar 2017 05:28:27 +0000 (13:28 +0800)]
btrfs-progs: dump-tree: Fix duplicated output when using -t option

When using -t option to output trees not in root tree (chunk/root/log
root), then we output the tree twice.

Fix it

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: misc/019, use fssum
David Sterba [Wed, 15 Mar 2017 11:28:16 +0000 (12:28 +0100)]
btrfs-progs: tests: misc/019, use fssum

The fssum utility is now ready to use, test still passes.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: fssum, switch from MD5 to SHA256
David Sterba [Wed, 15 Mar 2017 11:21:50 +0000 (12:21 +0100)]
btrfs-progs: tests: fssum, switch from MD5 to SHA256

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: add SHA256
David Sterba [Tue, 14 Mar 2017 18:30:47 +0000 (19:30 +0100)]
btrfs-progs: tests: add SHA256

In order to drop dependency on SSL library to compute MD5 in fssum,
we'll use the reference implementation from RFC 6234.

The checksum is not in a cryptographically sensitive context, but we're
going to skip MD5 and SHA-1 anyway.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: fssum, update includes
David Sterba [Tue, 14 Mar 2017 18:09:29 +0000 (19:09 +0100)]
btrfs-progs: tests: fssum, update includes

Remove unneeded and add headers that define time_te and int types.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: fssum, drop *_SOURCE defines, already provided via config.h
David Sterba [Tue, 14 Mar 2017 18:04:51 +0000 (19:04 +0100)]
btrfs-progs: tests: fssum, drop *_SOURCE defines, already provided via config.h

In file included from /usr/include/stdio.h:27:0,
                 from kerncompat.h:22,
                 from tests/fssum.c:25:
/usr/include/features.h:148:3: warning: #warning "_BSD_SOURCE and _SVID_SOURCE are deprecated, use _DEFAULT_SOURCE" [-Wcpp]
 # warning "_BSD_SOURCE and _SVID_SOURCE are deprecated, use _DEFAULT_SOURCE"
   ^~~~~~~

We've solved that long time ago and config.h now provides the macros.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: fssum, use our endianity helper
David Sterba [Tue, 14 Mar 2017 18:01:35 +0000 (19:01 +0100)]
btrfs-progs: tests: fssum, use our endianity helper

The function htonll is not provided by the standard library and we can
replace it by our cpu-to-XX helpers. This switches the endianity of the
checksummed value to LE, but this is not a problem.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: fssum, drop unnecessray build defines
David Sterba [Tue, 14 Mar 2017 17:56:22 +0000 (18:56 +0100)]
btrfs-progs: tests: fssum, drop unnecessray build defines

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: add fssum utility
David Sterba [Tue, 14 Mar 2017 17:53:29 +0000 (18:53 +0100)]
btrfs-progs: tests: add fssum utility

Copy from fstests, originally from
git://git.kernel.org/pub/scm/linux/kernel/git/arne/far-progs.git

Needs libcrypto to link but this check is now missing in configure.

Signed-off-by: Jan Schmidt <list.btrfs@jan-o-sch.net>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: cleanup use of IMAGE and replace by helpers
David Sterba [Mon, 13 Mar 2017 17:07:28 +0000 (18:07 +0100)]
btrfs-progs: tests: cleanup use of IMAGE and replace by helpers

Nearly each use of IMAGE can be replaced by common helpers as there are
no specific requirements on the testing filesystem. There are still a
few left that need to be evaluated and converted eventually.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: mkfs/002 and 003: use TEST_DEV instead of IMAGE
David Sterba [Mon, 13 Mar 2017 16:52:20 +0000 (17:52 +0100)]
btrfs-progs: tests: mkfs/002 and 003: use TEST_DEV instead of IMAGE

As we don't know what's the TEST_DEV like, use the explicit mkfs limit.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: use TEST_DEV in cli-tests
David Sterba [Mon, 13 Mar 2017 16:48:55 +0000 (17:48 +0100)]
btrfs-progs: tests: use TEST_DEV in cli-tests

IMAGE points to the fallback image, tests should use TEST_DEV.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: docs: add preliminary do's & dont's for tests
David Sterba [Mon, 13 Mar 2017 16:43:05 +0000 (17:43 +0100)]
btrfs-progs: docs: add preliminary do's & dont's for tests

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: do test build of library-test.static
David Sterba [Mon, 13 Mar 2017 16:15:24 +0000 (17:15 +0100)]
btrfs-progs: tests: do test build of library-test.static

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: library-test: update test to build outside of topdir
David Sterba [Mon, 13 Mar 2017 16:01:28 +0000 (17:01 +0100)]
btrfs-progs: library-test: update test to build outside of topdir

We need to build outside of the topdir so we can use the "btrfs/" prefix
for includes and not accidentally include other files.

Make magic is simple:
- build dependencies inside TOPDIR
- build inside temporary directory, link back to TOPDIR
- library-test.o not built anymore obviously

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: library-test: update header inclusion
David Sterba [Mon, 13 Mar 2017 15:52:34 +0000 (16:52 +0100)]
btrfs-progs: library-test: update header inclusion

The library-test is supposed to compile and link as an external
program, so we should update the way the headers are included.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: library: distribute sizes.h, fix build
David Sterba [Mon, 13 Mar 2017 15:08:51 +0000 (16:08 +0100)]
btrfs-progs: library: distribute sizes.h, fix build

The patch "btrfs-progs: Introduce kernel sizes to cleanup large
intermediate number" (a2203246ae483d09) was taken from kernel but not
properly ported so the build breaks because the header linux/sizes.h is
not exported.

The build tests of library do not cover the case when the macro
BTRFS_FLAT_INCLUDES is not defined (ie. an external build).

Reported-by: Holger Hoffstätte <holger@applied-asynchrony.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: add missing limits headers
Baruch Siach [Sun, 12 Mar 2017 17:46:52 +0000 (19:46 +0200)]
btrfs-progs: add missing limits headers

Build under musl libc fails because of missing PATH_MAX and XATTR_NAME_MAX
macro declarations. Add the required headers.

Signed-off-by: Baruch Siach <baruch@tkos.co.il>
7 years agobtrfs-progs: convert: Rework rollback
Qu Wenruo [Thu, 23 Feb 2017 08:21:14 +0000 (16:21 +0800)]
btrfs-progs: convert: Rework rollback

Rework rollback to a more easy to understand way.

New convert behavior makes us to have a more flex chunk layout, which
only data chunk containing old fs data will be at the same physical
location, while new chunks (data/meta/sys) can be mapped anywhere else.

This behavior makes old rollback behavior can't handle it.
As old behavior assumes all data/meta is mapped in a large chunk, which is
mapped 1:1 on disk.

So rework rollback to handle new convert behavior, enhance the check by
only checking all file extents of convert image, only to check if these
file extents and therir chunks are mapped 1:1.

This new rollback check behavior can handle both new and old convert
behavior, as the new behavior is a superset of old behavior.

Further more, introduce a simple rollback mechanisim:
1) Read reserved data (offset = file offset) from convert image
2) Write reserved data into disk (offset = physical offset)

Since old fs image is a valid fs, and we only need to rollback
superblocks (btrfs reserved ranges), then we just read out data in
reserved range, and write it back.

Due to the fact that all other file extents of converted image is mapped
1:1 on disk, we put the missing piece back, then the fs is as good as
old one.

Then what we do in btrfs is just another dream.

With this new rollback mechanisim, we can open btrfs read-only, so we
won't cause any damage to current btrfs, until the final piece (0~1M,
containing 1st super block) is put back.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
[ port to v4.10 ]
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: convert: Introduce function to check if convert image is able to be...
Qu Wenruo [Thu, 23 Feb 2017 07:46:46 +0000 (15:46 +0800)]
btrfs-progs: convert: Introduce function to check if convert image is able to be rolled back

Introduce a function, check_convert_image() to check if that image is
rollback-able.

This means all file extents except one of the btrfs reserved ranges, must
be mapped 1:1 on disk.

1:1 mapped file extents must match the following conditions:

1) Their file_offset(key.offset) matches its disk_bytenr
2) The corresponding chunk must be mapped 1:1 on disk
   That's to say, it's a SINGLE chunk, and chunk logical matches with
   stripe physical.

Above 2 conditions ensured that file extent lies the exactly the same
position as in the old filesystem.

For data in reserved ranges of btrfs, they are relocated to new places,
and in that case, we use btrfs_read_file() to read out the content for
later rollback use.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
7 years agobtrfs-progs: convert: Introduce function to read out btrfs reserved range
Qu Wenruo [Thu, 23 Feb 2017 07:17:16 +0000 (15:17 +0800)]
btrfs-progs: convert: Introduce function to read out btrfs reserved range

Introduce a new function, read_reserved_ranges(), to allow later
rollback to use these data to do rollback.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
7 years agobtrfs-progs: file: Introduce function to read out file content
Qu Wenruo [Thu, 23 Feb 2017 05:53:25 +0000 (13:53 +0800)]
btrfs-progs: file: Introduce function to read out file content

Introduce a new function, btrfs_read_file(), to read out data of a file
inode.

This function will iterate through EXTENT_DATA items and handle
inline/prealloc/hole file extents.

Compression is not supported yet.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: convert: Use reserved ranges array to cleanup open code
Qu Wenruo [Thu, 23 Feb 2017 02:22:22 +0000 (10:22 +0800)]
btrfs-progs: convert: Use reserved ranges array to cleanup open code

Since we have reserved ranges array now, we can use them to skip all
these open codes.

And add some comment and asciidoc art for related part.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
[ port to v4.10 ]
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: convert: Introduce simple range structure for convert reserved ranges
Qu Wenruo [Wed, 22 Feb 2017 08:46:18 +0000 (16:46 +0800)]
btrfs-progs: convert: Introduce simple range structure for convert reserved ranges

Introduce a new strucutre, simple_range, to present one contingous
range.

Also, use such structure to define btrfs_reserved_ranges(), which
convert and rollback will use.

Suggested-by: David Sterba <dsterba@suse.com>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
[ split hunks to new file structure ]
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: convert: Add comment for the overall convert design
Qu Wenruo [Wed, 22 Feb 2017 07:50:32 +0000 (15:50 +0800)]
btrfs-progs: convert: Add comment for the overall convert design

Convert is now a little complex due to that fact we need to separate
metadata and data chunks for different profiles.

Add a comment with ascii art explaining the whole design and point
out the really complex part, so any newcomers interested in convert can
get a quick overview of it before digging into the hard to read code.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
[ wording and formatting adjustments ]
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: port 019-receive-clones-on-munted-subvol
David Sterba [Wed, 8 Mar 2017 14:51:57 +0000 (15:51 +0100)]
btrfs-progs: tests: port 019-receive-clones-on-munted-subvol

Port from fstests.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: correctly receive clones to mounted subvol
Benedikt Morbach [Wed, 22 Feb 2017 23:07:35 +0000 (00:07 +0100)]
btrfs-progs: tests: correctly receive clones to mounted subvol

clone needs to resolve the paths of the involved subvolumes in the target
fs from their UUIDs. When doing so it might need to strip the prefix
that is mounted as the root of the fs from those paths.

It didn't do so correctly when processing the source of "clone" commands

This is a regression test for
    btrfs-progs: receive: handle root subvol path in clone

Signed-off-by: Benedikt Morbach <benedikt.morbach@googlemail.com>
[ copied the fstests version, will be updated later ]
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: receive: handle root subvol path in clone
Benedikt Morbach [Wed, 22 Feb 2017 22:56:37 +0000 (23:56 +0100)]
btrfs-progs: receive: handle root subvol path in clone

testcase:
    # ro subvol /src/parent
    # rw subvol /src/foo
    clone /src/parent/file /src/foo/file
    subvol snapshot -r /src/foo /src/foo.snap

    # generates a "clone parent/file -> foo.snap/file" send command
    send -p /src/parent /src/foo.snap

    # target fs:
    #    dest/
    #        |--- parent/...
    # mounted with -o subvol=dest, such that "parent" is at <target>/parent
    receive <target>

result:
    ERROR: cannot open dest/parent/file: No such file or directory

expected:
    "dest/" get's stripped from the clone source path to get the actual
    path in the target fs, if reachable from the mount point/chroot.

    This is exactly what process_snapshot does, which gets called on
    _every_ incremental receive and I'm quite certain is correct in
    doing so

Signed-off-by: Benedikt Morbach <benedikt.morbach@googlemail.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: receive: better error reporting for snapshots
Benedikt Morbach [Wed, 22 Feb 2017 22:56:36 +0000 (23:56 +0100)]
btrfs-progs: receive: better error reporting for snapshots

Two fixes:

1)

Check that the parent subvol actually is reachable via our root path.
The previous code wouldn't catch

    parent subvol: foo/bar
    root path:     bar   (i.e. mounted with -o subvol=bar)

where the parent isn't reachable from the root path.
(but the original "strstr(parent, root_path) == NULL" check still doesn't hold)

Also check for the slash after "root_path", i.e. throw an error on

    parent subvol: foobar
    root path:     foo

2)

If the parent subvol is the one that is mounted we obviously can't
receive into it, as it has to be read-only by definition.

We'd get a rather cryptic:

    At subvol /tmp/test/dest.snap
    At snapshot dest.snap
    ERROR: creating snapshot / -> dest.snap failed: Invalid cross-device link

(not sure what it says if "/" isn't even a btrfs)

But with this we get

    At subvol /tmp/test/dest.snap
    At snapshot dest.snap
    ERROR: creating snapshot . -> dest.snap failed: Read-only file system

which is both more helpful and more correct.

Signed-off-by: Benedikt Morbach <benedikt.morbach@googlemail.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agoBtrfs progs v4.10
David Sterba [Wed, 8 Mar 2017 13:23:24 +0000 (14:23 +0100)]
Btrfs progs v4.10

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: update CHANGES for v4.10
David Sterba [Thu, 2 Mar 2017 13:36:25 +0000 (14:36 +0100)]
btrfs-progs: update CHANGES for v4.10

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: report I/O errors when closing the filesystem
Omar Sandoval [Fri, 3 Mar 2017 17:02:14 +0000 (09:02 -0800)]
btrfs-progs: report I/O errors when closing the filesystem

If the final fsync() on the Btrfs device fails, we just swallow the
error and don't alert the user in any way. This was uncovered by xfstest
generic/405, which checks that mkfs fails when it encounters EIO.

Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: use dd in 025-file-extents for creating files
David Sterba [Wed, 1 Mar 2017 16:16:46 +0000 (17:16 +0100)]
btrfs-progs: tests: use dd in 025-file-extents for creating files

Drop the dependency on xfs_io as it's not a standard tool, though it
provides convenience. We use a simple write here so dd can manage.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: fix permissions of temporary files in 025-file-extents
David Sterba [Wed, 1 Mar 2017 16:10:44 +0000 (17:10 +0100)]
btrfs-progs: tests: fix permissions of temporary files in 025-file-extents

If the tests are started from non-root user, the fallocate and xfs_io
fail. Use the root helper as a workaround, we'd should fix the perms
instead.

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: enhance return values of walk_down_tree_v2
Su Yue [Tue, 21 Feb 2017 08:34:38 +0000 (16:34 +0800)]
btrfs-progs: check: enhance return values of walk_down_tree_v2

In lowmem mode, 'walk_down_tree_v2' returns negative values wheather
the error is fatal or not. It causes the loop where 'walk_down_tree_v2'
is to break even the error is tolerated and then subsequent nodes process
will be skipped.

Fix it by redefining meanings of values 'walk_down_tree_v2' returns.
Do a similar fix for 'process_one_leaf_v2'.

Signed-off-by: Su Yue <suy.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: Fix lowmem mode override to allow it skip repair work
Lu Fengqi [Tue, 21 Feb 2017 08:34:37 +0000 (16:34 +0800)]
btrfs-progs: check: Fix lowmem mode override to allow it skip repair work

Current common.local doesn't handle lowmem mode well.
It passes "--mode=lowmem" alone with "--repair", making it unable to
check lowmem mode.

It's caused by the following bugs:

1) Wrong variable in test/common.local
   We should check TEST_ARGS_CHECK, not TEST_CHECK, which is not defined
   so we never return 1.

2) Wrong parameter passed to _cmd_spec() in test/common
   This prevents us from grepping the correct parameters.

Fix it.

Signed-off-by: Lu Fengqi <lufq.fnst@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: Add new test case for file extent false alerts
Qu Wenruo [Tue, 21 Feb 2017 08:34:36 +0000 (16:34 +0800)]
btrfs-progs: tests: Add new test case for file extent false alerts

Lowmem mode exposed several false alerts, all related to file extents
check.

1) Partly written prealloc extent
   Cause lowmem mode to report missing csum or prealloc extent should
   not have csum

2) Compressed inline extent
   Cause lowmem mode to find mismatch on inline len and item len.
   While no error message is output but exit silently.

Reported-by: Chris Murphy <chris@colorremedies.com>
Reported-by: Christoph Anton Mitterer <calestyo@scientia.net>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: Make fsck-test/013 compatible with lowmem mode
Qu Wenruo [Tue, 21 Feb 2017 08:34:35 +0000 (16:34 +0800)]
btrfs-progs: tests: Make fsck-test/013 compatible with lowmem mode

fsck-tests/013-extent-tree-rebuild uses "--init-extent-tree", which
implies "--repair".

But the test script doesn't specify "--repair" for lowmem mode test to
detect it.

Add it so lowmem mode test can be happy with it.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: Add test image for lowmem mode block group false alert
Qu Wenruo [Tue, 21 Feb 2017 08:34:34 +0000 (16:34 +0800)]
btrfs-progs: tests: Add test image for lowmem mode block group false alert

Add a minimal image which can reproduce the block group used space
false alert for lowmem mode fsck.

Reported-by: Christoph Anton Mitterer <calestyo@scientia.net>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: Move fsck-tests/015 to fuzz tests
Qu Wenruo [Tue, 21 Feb 2017 08:34:33 +0000 (16:34 +0800)]
btrfs-progs: tests: Move fsck-tests/015 to fuzz tests

The test case fsck-tests/015-check-bad-memory-access can't be repair by
btrfs check, and it's a fortunate bug makes original mode to forget the
error code from extent tree, making original mode pass it.

So fuzz-tests is more suitable for it.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: lowmem: Fix silent error if first inode item missing
Qu Wenruo [Tue, 21 Feb 2017 08:34:32 +0000 (16:34 +0800)]
btrfs-progs: check: lowmem: Fix silent error if first inode item missing

If first inode item is missing, lowmem check will detect it but does not
output any error message.

Add error message for it.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: lowmem: Fix false alert on inline compressed extent
Qu Wenruo [Tue, 21 Feb 2017 08:34:31 +0000 (16:34 +0800)]
btrfs-progs: check: lowmem: Fix false alert on inline compressed extent

Old lowmem check doesn't check if the inline extent is compressed and
always checks extent numbytes against inline item size.

And when it finds the extent numbytes mismatch with inline item size it
doesn't output any error message, just return error silently, making it
quite hard to debug.

Fix it by only checking extent numbytes against inline item size when
the extent is not compressed, and output error message.

Reported-by: Christoph Anton Mitterer <calestyo@scientia.net>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: lowmem: Fix extent item size false alert
Qu Wenruo [Tue, 21 Feb 2017 08:34:30 +0000 (16:34 +0800)]
btrfs-progs: check: lowmem: Fix extent item size false alert

If one extent item has no inline ref, btrfs lowmem mode check can give
false alert without outputting any error message.

The problem is lowmem mode always assumes that extent item has inline
refs, and when it encounters such case it flags the extent item has
wrong size, but doesn't output the error message.

Although we already have such image submitted, at the commit time due to
another bug in cmds-check return value, it doesn't detect it until that
bug is fixed.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: lowmem: Fix false alert in checking data extent csums
Qu Wenruo [Tue, 21 Feb 2017 08:34:29 +0000 (16:34 +0800)]
btrfs-progs: check: lowmem: Fix false alert in checking data extent csums

Btrfs lowmem check can report false csum error like:
ERROR: root 5 EXTENT_DATA[257 0] datasum missing
ERROR: root 5 EXTENT_DATA[257 4096] prealloc shouldn't have datasum

This is because lowmem check code always compare the found csum size
with the whole extent which data extents points to.

Normally it's OK, but when prealloc extent is written, or reflink is
done, data extent can points to part of a larger extent, making the csum
check wrong.

To fix it, the csum check part is modified to handle plain and
compressed extents in different ways:

1) Plain extent
   Only search csums for the range it refers to.
   So the search range is from (disk_bytenr + extent_offset) and search
   length is (extent_num_bytes)

2) Compressed extent
   Search the whole extent.
   Search range is from (disk_bytner) and search length is
   (disk_num_bytes)

Reported-by: Chris Murphy <chris@colorremedies.com>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: Output verbose error when fsck found a bug in any tree
Qu Wenruo [Tue, 21 Feb 2017 08:34:28 +0000 (16:34 +0800)]
btrfs-progs: check: Output verbose error when fsck found a bug in any tree

Although we output error like "errors found in extent allocation tree or
chunk allocation", but we lacks such output for other trees, but leaving
the final "found error is %d" to catch the last return value(and
sometime it's cleared)

This patch adds extra error message for top level error path, and modify
the last "found error is %d" to "error(s) found" or "no error found".

Cc: Christoph Anton Mitterer <calestyo@scientia.net>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: check: lowmem: Fix several bugs related to afterward search
Qu Wenruo [Tue, 21 Feb 2017 08:34:27 +0000 (16:34 +0800)]
btrfs-progs: check: lowmem: Fix several bugs related to afterward search

Since btrfs_search_slot() can point to the slot which is beyond the
leaves' capacity, we should pay extra attention when doing afterward
search.

While for lowmem check, several places uses afterward search:
1) Block group item used space check
2) Device item used space check
3) Data extent backref check.

In the following case for block group item check, btrfs lowmem mode
check will skip the block group and report false alert:

leaf 29405184 items 37 free space 1273 generation 11 owner 2
...
        item 36 key (77594624 EXTENT_ITEM 2097152)
                extent refs 1 gen 8 flags DATA
                extent data backref root 5 objectid 265 offset 0 count 1
leaf 29409280 items 43 free space 670 generation 11 owner 2
        item 0 key (96468992 EXTENT_ITEM 2097152)
                extent refs 1 gen 8 flags DATA
                extent data backref root 5 objectid 274 offset 0 count 1
        item 1 key (96468992 BLOCK_GROUP_ITEM 33554432)
                block group used 2265088 chunk_objectid 256 flags DATA

When checking block group item, we will search key (96468992 0 0) to
start from the first item in the block group.

While search_slot() will point to leaf 29405184, slot 37 which is beyond
leaf capacity.

And when reading key from slot 37, uninitialized data can be read out
and cause us to exit block group item check, leading to false alert.

Fix it by checking path.slot[0] before reading out the key.

Reported-by: Christoph Anton Mitterer <calestyo@scientia.net>
Reported-by: Chris Murphy <chris@colorremedies.com>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: docs: fix many typos, plus three edits for clarity
Nicholas D Steeves [Tue, 21 Feb 2017 23:14:38 +0000 (18:14 -0500)]
btrfs-progs: docs: fix many typos, plus three edits for clarity

Signed-off-by: Nicholas D Steeves <nsteeves@gmail.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: send-dump: add missing newlines
Benedikt Morbach [Wed, 22 Feb 2017 21:32:46 +0000 (22:32 +0100)]
btrfs-progs: send-dump: add missing newlines

make sure to include newlines after commands that have only one
argument, such as 'unlink' or 'mkfile'

changes

    unlink          ./baz.0/file_autimes          ./baz.0/                        atime=2017-02-22T11:59:16+0100 mtime=2017-02-22T11:59:16+0100 ctime=2017-02-22T11:59:16+0100
    truncate        ./baz.0/file_a                  size=131072
    chmod           ./baz.0/file_a                  mode=644
    utimes          ./baz.0/file_a                  atime=2017-02-22T11:59:16+0100 mtime=2017-02-22T11:59:16+0100 ctime=2017-02-22T11:59:16+0100
    mkfile          ./baz.0/o258-11-0rename          ./baz.0/o258-11-0               dest=./baz.0/file_b
    utimes          ./baz.0/                        atime=2017-02-22T11:59:16+0100 mtime=2017-02-22T11:59:16+0100 ctime=2017-02-22T11:59:16+0100

to

    unlink          ./baz.0/file_a
    utimes          ./baz.0/                        atime=2017-02-22T11:59:16+0100 mtime=2017-02-22T11:59:16+0100 ctime=2017-02-22T11:59:16+0100
    truncate        ./baz.0/file_a                  size=131072
    chmod           ./baz.0/file_a                  mode=644
    utimes          ./baz.0/file_a                  atime=2017-02-22T11:59:16+0100 mtime=2017-02-22T11:59:16+0100 ctime=2017-02-22T11:59:16+0100
    mkfile          ./baz.0/o258-11-0
    rename          ./baz.0/o258-11-0               dest=./baz.0/file_b
    utimes          ./baz.0/                        atime=2017-02-22T11:59:16+0100 mtime=2017-02-22T11:59:16+0100 ctime=2017-02-22T11:59:16+0100

Signed-off-by: Benedikt Morbach <benedikt.morbach@googlemail.com>
Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: docs: fix typo in btrfs-quota
Joey Sabey [Wed, 22 Feb 2017 09:08:17 +0000 (09:08 +0000)]
btrfs-progs: docs: fix typo in btrfs-quota

qgruops -> qgroups

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: docs: fix typo in btrfs-quota
Joey Sabey [Wed, 22 Feb 2017 09:04:19 +0000 (09:04 +0000)]
btrfs-progs: docs: fix typo in btrfs-quota

reinstall ation -> reinstallation

Signed-off-by: David Sterba <dsterba@suse.com>
7 years agobtrfs-progs: tests: fix library-test rules and deps
David Sterba [Mon, 20 Feb 2017 13:58:57 +0000 (14:58 +0100)]
btrfs-progs: tests: fix library-test rules and deps

The symbol __error was undefined and the rule did not use the
dependencies from the rule definiton, like the rest.

Signed-off-by: David Sterba <dsterba@suse.com>