As Jungyeon Reported in bugzilla:
https://bugzilla.kernel.org/show_bug.cgi?id=203233
- Reproduces
gcc poc_13.c
./run.sh f2fs
- Kernel messages
F2FS-fs (sdb): Bitmap was wrongly set, blk:4608
kernel BUG at fs/f2fs/segment.c:2133!
RIP: 0010:update_sit_entry+0x35d/0x3e0
Call Trace:
f2fs_allocate_data_block+0x16c/0x5a0
do_write_page+0x57/0x100
f2fs_do_write_node_page+0x33/0xa0
__write_node_page+0x270/0x4e0
f2fs_sync_node_pages+0x5df/0x670
f2fs_write_checkpoint+0x364/0x13a0
f2fs_sync_fs+0xa3/0x130
f2fs_do_sync_file+0x1a6/0x810
do_fsync+0x33/0x60
__x64_sys_fsync+0xb/0x10
do_syscall_64+0x43/0x110
entry_SYSCALL_64_after_hwframe+0x44/0xa9
The testcase fails because that, in fuzzed image, current segment was
allocated with LFS type, its .next_blkoff should point to an unused
block address, but actually, its bitmap shows it's not. So during
allocation, f2fs crash when setting bitmap.
Introducing sanity_check_curseg() to check such inconsistence of
current in-used segment.
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
ext4_break_layouts() may fail e.g. due to a signal being delivered.
Thus we need to handle its failure gracefully and not by taking the
filesystem down. Currently ext4_break_layouts() failure is rare but it
may become more common once RDMA uses layout leases for handling
long-term page pins for DAX mappings.
To handle the failure we need to move ext4_break_layouts() earlier
during setattr handling before we do hard to undo changes such as
modifying inode size. To be able to do that we also have to move some
other checks which are better done without holding i_mmap_sem earlier.
Reported-and-tested-by: Ira Weiny <ira.weiny@intel.com>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
There is a missing brelse of bitmap_bh in an error
path of ext2_new_inode().
Signed-off-by: Chengguang Xu <cgxu519@zoho.com.cn>
Signed-off-by: Jan Kara <jack@suse.cz>
In cifs_read_allocate_pages, in case of ENOMEM, we go through
whole rdata->pages array but we have failed the allocation before
nr_pages, therefore we may end up calling put_page with NULL
pointer, causing oops
Signed-off-by: Roberto Bergantinos Corpas <rbergant@redhat.com>
Acked-by: Pavel Shilovsky <pshilov@microsoft.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
CC: Stable <stable@vger.kernel.org>
Overlapping overlay layers are not supported and can cause unexpected
behavior, but overlayfs does not currently check or warn about these
configurations.
User is not supposed to specify the same directory for upper and
lower dirs or for different lower layers and user is not supposed to
specify directories that are descendants of each other for overlay
layers, but that is exactly what this zysbot repro did:
https://syzkaller.appspot.com/x/repro.syz?x=12c7a94f400000
Moving layer root directories into other layers while overlayfs
is mounted could also result in unexpected behavior.
This commit places "traps" in the overlay inode hash table.
Those traps are dummy overlay inodes that are hashed by the layers
root inodes.
On mount, the hash table trap entries are used to verify that overlay
layers are not overlapping. While at it, we also verify that overlay
layers are not overlapping with directories "in-use" by other overlay
instances as upperdir/workdir.
On lookup, the trap entries are used to verify that overlay layers
root inodes have not been moved into other layers after mount.
Some examples:
$ ./run --ov --samefs -s
...
( mkdir -p base/upper/0/u base/upper/0/w base/lower lower upper mnt
mount -o bind base/lower lower
mount -o bind base/upper upper
mount -t overlay none mnt ...
-o lowerdir=lower,upperdir=upper/0/u,workdir=upper/0/w)
$ umount mnt
$ mount -t overlay none mnt ...
-o lowerdir=base,upperdir=upper/0/u,workdir=upper/0/w
[ 94.434900] overlayfs: overlapping upperdir path
mount: mount overlay on mnt failed: Too many levels of symbolic links
$ mount -t overlay none mnt ...
-o lowerdir=upper/0/u,upperdir=upper/0/u,workdir=upper/0/w
[ 151.350132] overlayfs: conflicting lowerdir path
mount: none is already mounted or mnt busy
$ mount -t overlay none mnt ...
-o lowerdir=lower:lower/a,upperdir=upper/0/u,workdir=upper/0/w
[ 201.205045] overlayfs: overlapping lowerdir path
mount: mount overlay on mnt failed: Too many levels of symbolic links
$ mount -t overlay none mnt ...
-o lowerdir=lower,upperdir=upper/0/u,workdir=upper/0/w
$ mv base/upper/0/ base/lower/
$ find mnt/0
mnt/0
mnt/0/w
find: 'mnt/0/w/work': Too many levels of symbolic links
find: 'mnt/0/u': Too many levels of symbolic links
Reported-by: syzbot+9c69c282adc4edd2b540@syzkaller.appspotmail.com
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
In flush_cache_ent(), 'ce->ce_path' is allocated by kstrdup_const().
It should be freed by kfree_const(), rather than kfree().
Signed-off-by: Gen Zhang <blackgod016574@gmail.com>
Reviewed-by: Paulo Alcantara <palcantara@suse.de>
Signed-off-by: Steve French <stfrench@microsoft.com>
The 2nd buffer could be NULL even if iov_len is not zero. This can
trigger a panic when handling symlinks. It's easy to reproduce with
LTP fs_racer scripts[1] which are randomly craete/delete/link files
and dirs. Fix this panic by checking if the 2nd buffer is padding
before kfree, like what we do in SMB2_open_free.
[1] https://github.com/linux-test-project/ltp/tree/master/testcases/kernel/fs/racer
Fixes: 2c87d6a94d ("cifs: Allocate memory for all iovs in smb2_ioctl")
Signed-off-by: Murphy Zhou <jencce.kernel@gmail.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
Reviewed-by: Ronnie sahlberg <lsahlber@redhat.com>
Currently in the case where SMB2_ioctl returns the -EOPNOTSUPP error
there is a memory leak of pneg_inbuf. Fix this by returning via
the out_free_inbuf exit path that will perform the relevant kfree.
Addresses-Coverity: ("Resource leak")
Fixes: 969ae8e8d4 ("cifs: Accept validate negotiate if server return NT_STATUS_NOT_SUPPORTED")
CC: Stable <stable@vger.kernel.org> # v5.1+
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
The directory may have been removed when entering
fscrypt_ioctl_set_policy(). If so, the empty_dir() check will return
error for ext4 file system.
ext4_rmdir() sets i_size = 0, then ext4_empty_dir() reports an error
because 'inode->i_size < EXT4_DIR_REC_LEN(1) + EXT4_DIR_REC_LEN(2)'. If
the fs is mounted with errors=panic, it will trigger a panic issue.
Add the check IS_DEADDIR() to fix this problem.
Fixes: 9bd8212f98 ("ext4 crypto: add encryption policy and password salt support")
Cc: <stable@vger.kernel.org> # v4.1+
Signed-off-by: Hongjie Fang <hongjiefang@asrmicro.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
As an optimization, don't encrypt blocks fully beyond i_size, since
those definitely won't need to be written out. Also add a comment.
This is in preparation for allowing encryption on ext4 filesystems with
blocksize != PAGE_SIZE.
This is based on work by Chandan Rajendra.
Reviewed-by: Chandan Rajendra <chandan@linux.ibm.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
In __ext4_block_zero_page_range(), only decrypt the block that actually
needs to be decrypted, rather than assuming blocksize == PAGE_SIZE and
decrypting the whole page.
This is in preparation for allowing encryption on ext4 filesystems with
blocksize != PAGE_SIZE.
Signed-off-by: Chandan Rajendra <chandan@linux.ibm.com>
(EB: rebase onto previous changes, improve the commit message, and use
bh_offset())
Signed-off-by: Eric Biggers <ebiggers@google.com>
In ext4_block_write_begin(), only decrypt the blocks that actually need
to be decrypted (up to two blocks which intersect the boundaries of the
region that will be written to), rather than assuming blocksize ==
PAGE_SIZE and decrypting the whole page.
This is in preparation for allowing encryption on ext4 filesystems with
blocksize != PAGE_SIZE.
Signed-off-by: Chandan Rajendra <chandan@linux.ibm.com>
(EB: rebase onto previous changes, improve the commit message,
and move the check for encrypted inode)
Signed-off-by: Eric Biggers <ebiggers@google.com>
If decryption fails, ext4_block_write_begin() can return with the page's
buffer_head marked with the BH_Uptodate flag. This commit clears the
BH_Uptodate flag in such cases.
Signed-off-by: Chandan Rajendra <chandan@linux.ibm.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
In __fscrypt_decrypt_bio(), only decrypt the blocks that actually
comprise the bio, rather than assuming blocksize == PAGE_SIZE and
decrypting the entirety of every page used in the bio.
This is in preparation for allowing encryption on ext4 filesystems with
blocksize != PAGE_SIZE.
This is based on work by Chandan Rajendra.
Reviewed-by: Chandan Rajendra <chandan@linux.ibm.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Rename fscrypt_decrypt_page() to fscrypt_decrypt_pagecache_blocks() and
redefine its behavior to decrypt all filesystem blocks in the given
region of the given page, rather than assuming that the region consists
of just one filesystem block. Also remove the 'inode' and 'lblk_num'
parameters, since they can be retrieved from the page as it's already
assumed to be a pagecache page.
This is in preparation for allowing encryption on ext4 filesystems with
blocksize != PAGE_SIZE.
This is based on work by Chandan Rajendra.
Reviewed-by: Chandan Rajendra <chandan@linux.ibm.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Currently fscrypt_decrypt_page() does one of two logically distinct
things depending on whether FS_CFLG_OWN_PAGES is set in the filesystem's
fscrypt_operations: decrypt a pagecache page in-place, or decrypt a
filesystem block in-place in any page. Currently these happen to share
the same implementation, but this conflates the notion of blocks and
pages. It also makes it so that all callers have to provide inode and
lblk_num, when fscrypt could determine these itself for pagecache pages.
Therefore, move the FS_CFLG_OWN_PAGES behavior into a new function
fscrypt_decrypt_block_inplace(). This mirrors
fscrypt_encrypt_block_inplace().
This is in preparation for allowing encryption on ext4 filesystems with
blocksize != PAGE_SIZE.
Reviewed-by: Chandan Rajendra <chandan@linux.ibm.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Adjust fscrypt_zeroout_range() to encrypt a block at a time rather than
a page at a time, so that it works when blocksize < PAGE_SIZE.
This isn't optimized for performance, but then again this function
already wasn't optimized for performance. As a future optimization, we
could submit much larger bios here.
This is in preparation for allowing encryption on ext4 filesystems with
blocksize != PAGE_SIZE.
This is based on work by Chandan Rajendra.
Reviewed-by: Chandan Rajendra <chandan@linux.ibm.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Rename fscrypt_encrypt_page() to fscrypt_encrypt_pagecache_blocks() and
redefine its behavior to encrypt all filesystem blocks from the given
region of the given page, rather than assuming that the region consists
of just one filesystem block. Also remove the 'inode' and 'lblk_num'
parameters, since they can be retrieved from the page as it's already
assumed to be a pagecache page.
This is in preparation for allowing encryption on ext4 filesystems with
blocksize != PAGE_SIZE.
This is based on work by Chandan Rajendra.
Reviewed-by: Chandan Rajendra <chandan@linux.ibm.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
fscrypt_encrypt_page() behaves very differently depending on whether the
filesystem set FS_CFLG_OWN_PAGES in its fscrypt_operations. This makes
the function difficult to understand and document. It also makes it so
that all callers have to provide inode and lblk_num, when fscrypt could
determine these itself for pagecache pages.
Therefore, move the FS_CFLG_OWN_PAGES behavior into a new function
fscrypt_encrypt_block_inplace().
This is in preparation for allowing encryption on ext4 filesystems with
blocksize != PAGE_SIZE.
Reviewed-by: Chandan Rajendra <chandan@linux.ibm.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Replace some BUG_ON()s with WARN_ON_ONCE() and returning an error code,
and move the check for len divisible by FS_CRYPTO_BLOCK_SIZE into
fscrypt_crypt_block() so that it's done for both encryption and
decryption, not just encryption.
Reviewed-by: Chandan Rajendra <chandan@linux.ibm.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
fscrypt_do_page_crypto() only does a single encryption or decryption
operation, with a single logical block number (single IV). So it
actually operates on a filesystem block, not a "page" per se. To
reflect this, rename it to fscrypt_crypt_block().
Reviewed-by: Chandan Rajendra <chandan@linux.ibm.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Now that fscrypt_ctx is not used for writes, remove the 'w' fields.
Reviewed-by: Chandan Rajendra <chandan@linux.ibm.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Currently, bounce page handling for writes to encrypted files is
unnecessarily complicated. A fscrypt_ctx is allocated along with each
bounce page, page_private(bounce_page) points to this fscrypt_ctx, and
fscrypt_ctx::w::control_page points to the original pagecache page.
However, because writes don't use the fscrypt_ctx for anything else,
there's no reason why page_private(bounce_page) can't just point to the
original pagecache page directly.
Therefore, this patch makes this change. In the process, it also cleans
up the API exposed to filesystems that allows testing whether a page is
a bounce page, getting the pagecache page from a bounce page, and
freeing a bounce page.
Reviewed-by: Chandan Rajendra <chandan@linux.ibm.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
When syncing the log, the final phase of a fsync operation, we need to
either create a log root's item or update the existing item in the log
tree of log roots, and that depends on the current value of the log
root's log_transid - if it's 1 we need to create the log root item,
otherwise it must exist already and we update it. Since there is no
synchronization between updating the log_transid and checking it for
deciding whether the log root's item needs to be created or updated, we
end up with a tiny race window that results in attempts to update the
item to fail because the item was not yet created:
CPU 1 CPU 2
btrfs_sync_log()
lock root->log_mutex
set log root's log_transid to 1
unlock root->log_mutex
btrfs_sync_log()
lock root->log_mutex
sets log root's
log_transid to 2
unlock root->log_mutex
update_log_root()
sees log root's log_transid
with a value of 2
calls btrfs_update_root(),
which fails with -EUCLEAN
and causes transaction abort
Until recently the race lead to a BUG_ON at btrfs_update_root(), but after
the recent commit 7ac1e464c4 ("btrfs: Don't panic when we can't find a
root key") we just abort the current transaction.
A sample trace of the BUG_ON() on a SLE12 kernel:
------------[ cut here ]------------
kernel BUG at ../fs/btrfs/root-tree.c:157!
Oops: Exception in kernel mode, sig: 5 [#1]
SMP NR_CPUS=2048 NUMA pSeries
(...)
Supported: Yes, External
CPU: 78 PID: 76303 Comm: rtas_errd Tainted: G X 4.4.156-94.57-default #1
task: c00000ffa906d010 ti: c00000ff42b08000 task.ti: c00000ff42b08000
NIP: d000000036ae5cdc LR: d000000036ae5cd8 CTR: 0000000000000000
REGS: c00000ff42b0b860 TRAP: 0700 Tainted: G X (4.4.156-94.57-default)
MSR: 8000000002029033 <SF,VEC,EE,ME,IR,DR,RI,LE> CR: 22444484 XER: 20000000
CFAR: d000000036aba66c SOFTE: 1
GPR00: d000000036ae5cd8 c00000ff42b0bae0 d000000036bda220 0000000000000054
GPR04: 0000000000000001 0000000000000000 c00007ffff8d37c8 0000000000000000
GPR08: c000000000e19c00 0000000000000000 0000000000000000 3736343438312079
GPR12: 3930373337303434 c000000007a3a800 00000000007fffff 0000000000000023
GPR16: c00000ffa9d26028 c00000ffa9d261f8 0000000000000010 c00000ffa9d2ab28
GPR20: c00000ff42b0bc48 0000000000000001 c00000ff9f0d9888 0000000000000001
GPR24: c00000ffa9d26000 c00000ffa9d261e8 c00000ffa9d2a800 c00000ff9f0d9888
GPR28: c00000ffa9d26028 c00000ffa9d2aa98 0000000000000001 c00000ffa98f5b20
NIP [d000000036ae5cdc] btrfs_update_root+0x25c/0x4e0 [btrfs]
LR [d000000036ae5cd8] btrfs_update_root+0x258/0x4e0 [btrfs]
Call Trace:
[c00000ff42b0bae0] [d000000036ae5cd8] btrfs_update_root+0x258/0x4e0 [btrfs] (unreliable)
[c00000ff42b0bba0] [d000000036b53610] btrfs_sync_log+0x2d0/0xc60 [btrfs]
[c00000ff42b0bce0] [d000000036b1785c] btrfs_sync_file+0x44c/0x4e0 [btrfs]
[c00000ff42b0bd80] [c00000000032e300] vfs_fsync_range+0x70/0x120
[c00000ff42b0bdd0] [c00000000032e44c] do_fsync+0x5c/0xb0
[c00000ff42b0be10] [c00000000032e8dc] SyS_fdatasync+0x2c/0x40
[c00000ff42b0be30] [c000000000009488] system_call+0x3c/0x100
Instruction dump:
7f43d378 4bffebb9 60000000 88d90008 3d220000 e8b90000 3b390009 e87a01f0
e8898e08 e8f90000 4bfd48e5 60000000 <0fe00000> e95b0060 39200004 394a0ea0
---[ end trace 8f2dc8f919cabab8 ]---
So fix this by doing the check of log_transid and updating or creating the
log root's item while holding the root's log_mutex.
Fixes: 7237f18336 ("Btrfs: fix tree logs parallel sync")
CC: stable@vger.kernel.org # 4.4+
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When replaying a log that contains a new file or directory name that needs
to be added to its parent directory, we end up updating the mtime and the
ctime of the parent directory to the current time after we have set their
values to the correct ones (set at fsync time), efectivelly losing them.
Sample reproducer:
$ mkfs.btrfs -f /dev/sdb
$ mount /dev/sdb /mnt
$ mkdir /mnt/dir
$ touch /mnt/dir/file
# fsync of the directory is optional, not needed
$ xfs_io -c fsync /mnt/dir
$ xfs_io -c fsync /mnt/dir/file
$ stat -c %Y /mnt/dir
1557856079
<power failure>
$ sleep 3
$ mount /dev/sdb /mnt
$ stat -c %Y /mnt/dir
1557856082
--> should have been 1557856079, the mtime is updated to the current
time when replaying the log
Fix this by not updating the mtime and ctime to the current time at
btrfs_add_link() when we are replaying a log tree.
This could be triggered by my recent fsync fuzz tester for fstests, for
which an fstests patch exists titled "fstests: generic, fsync fuzz tester
with fsstress".
Fixes: e02119d5a7 ("Btrfs: Add a write ahead tree log to optimize synchronous operations")
CC: stable@vger.kernel.org # 4.4+
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
While logging an inode we follow its ancestors and for each one we mark
it as logged in the current transaction, even if we have not logged it.
As a consequence if we change an attribute of an ancestor, such as the
UID or GID for example, and then explicitly fsync it, we end up not
logging the inode at all despite returning success to user space, which
results in the attribute being lost if a power failure happens after
the fsync.
Sample reproducer:
$ mkfs.btrfs -f /dev/sdb
$ mount /dev/sdb /mnt
$ mkdir /mnt/dir
$ chown 6007:6007 /mnt/dir
$ sync
$ chown 9003:9003 /mnt/dir
$ touch /mnt/dir/file
$ xfs_io -c fsync /mnt/dir/file
# fsync our directory after fsync'ing the new file, should persist the
# new values for the uid and gid.
$ xfs_io -c fsync /mnt/dir
<power failure>
$ mount /dev/sdb /mnt
$ stat -c %u:%g /mnt/dir
6007:6007
--> should be 9003:9003, the uid and gid were not persisted, despite
the explicit fsync on the directory prior to the power failure
Fix this by not updating the logged_trans field of ancestor inodes when
logging an inode, since we have not logged them. Let only future calls to
btrfs_log_inode() to mark inodes as logged.
This could be triggered by my recent fsync fuzz tester for fstests, for
which an fstests patch exists titled "fstests: generic, fsync fuzz tester
with fsstress".
Fixes: 12fcfd22fe ("Btrfs: tree logging unlink/rename fixes")
CC: stable@vger.kernel.org # 4.4+
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
When mounting a fs with reloc tree and has qgroup enabled, it can cause
NULL pointer dereference at mount time:
BUG: kernel NULL pointer dereference, address: 00000000000000a8
#PF: supervisor read access in kernel mode
#PF: error_code(0x0000) - not-present page
PGD 0 P4D 0
Oops: 0000 [#1] PREEMPT SMP NOPTI
RIP: 0010:btrfs_qgroup_add_swapped_blocks+0x186/0x300 [btrfs]
Call Trace:
replace_path.isra.23+0x685/0x900 [btrfs]
merge_reloc_root+0x26e/0x5f0 [btrfs]
merge_reloc_roots+0x10a/0x1a0 [btrfs]
btrfs_recover_relocation+0x3cd/0x420 [btrfs]
open_ctree+0x1bc8/0x1ed0 [btrfs]
btrfs_mount_root+0x544/0x680 [btrfs]
legacy_get_tree+0x34/0x60
vfs_get_tree+0x2d/0xf0
fc_mount+0x12/0x40
vfs_kern_mount.part.12+0x61/0xa0
vfs_kern_mount+0x13/0x20
btrfs_mount+0x16f/0x860 [btrfs]
legacy_get_tree+0x34/0x60
vfs_get_tree+0x2d/0xf0
do_mount+0x81f/0xac0
ksys_mount+0xbf/0xe0
__x64_sys_mount+0x25/0x30
do_syscall_64+0x65/0x240
entry_SYSCALL_64_after_hwframe+0x49/0xbe
[CAUSE]
In btrfs_recover_relocation(), we don't have enough info to determine
which block group we're relocating, but only to merge existing reloc
trees.
Thus in btrfs_recover_relocation(), rc->block_group is NULL.
btrfs_qgroup_add_swapped_blocks() hasn't taken this into consideration,
and causes a NULL pointer dereference.
The bug is introduced by commit 3d0174f78e ("btrfs: qgroup: Only trace
data extents in leaves if we're relocating data block group"), and
later qgroup refactoring still keeps this optimization.
[FIX]
Thankfully in the context of btrfs_recover_relocation(), there is no
other progress can modify tree blocks, thus those swapped tree blocks
pair will never affect qgroup numbers, no matter whatever we set for
block->trace_leaf.
So we only need to check if @bg is NULL before accessing @bg->flags.
Reported-by: Juan Erbes <jerbes@gmail.com>
Link: https://bugzilla.opensuse.org/show_bug.cgi?id=1134806
Fixes: 3d0174f78e ("btrfs: qgroup: Only trace data extents in leaves if we're relocating data block group")
CC: stable@vger.kernel.org # 4.20+
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
When a fs has orphan reloc tree along with unfinished balance:
...
item 16 key (TREE_RELOC ROOT_ITEM FS_TREE) itemoff 12090 itemsize 439
generation 12 root_dirid 256 bytenr 300400640 level 1 refs 0 <<<
lastsnap 8 byte_limit 0 bytes_used 1359872 flags 0x0(none)
uuid 7c48d938-33a3-4aae-ab19-6e5c9d406e46
item 17 key (BALANCE TEMPORARY_ITEM 0) itemoff 11642 itemsize 448
temporary item objectid BALANCE offset 0
balance status flags 14
Then at mount time, we can hit the following kernel BUG_ON():
BTRFS info (device dm-3): relocating block group 298844160 flags metadata|dup
------------[ cut here ]------------
kernel BUG at fs/btrfs/relocation.c:1413!
invalid opcode: 0000 [#1] PREEMPT SMP NOPTI
CPU: 1 PID: 897 Comm: btrfs-balance Tainted: G O 5.2.0-rc1-custom #15
RIP: 0010:create_reloc_root+0x1eb/0x200 [btrfs]
Call Trace:
btrfs_init_reloc_root+0x96/0xb0 [btrfs]
record_root_in_trans+0xb2/0xe0 [btrfs]
btrfs_record_root_in_trans+0x55/0x70 [btrfs]
select_reloc_root+0x7e/0x230 [btrfs]
do_relocation+0xc4/0x620 [btrfs]
relocate_tree_blocks+0x592/0x6a0 [btrfs]
relocate_block_group+0x47b/0x5d0 [btrfs]
btrfs_relocate_block_group+0x183/0x2f0 [btrfs]
btrfs_relocate_chunk+0x4e/0xe0 [btrfs]
btrfs_balance+0x864/0xfa0 [btrfs]
balance_kthread+0x3b/0x50 [btrfs]
kthread+0x123/0x140
ret_from_fork+0x27/0x50
[CAUSE]
In btrfs, reloc trees are used to record swapped tree blocks during
balance.
Reloc tree either get merged (replace old tree blocks of its parent
subvolume) in next transaction if its ref is 1 (fresh).
Or is already merged and will be cleaned up if its ref is 0 (orphan).
After commit d2311e6985 ("btrfs: relocation: Delay reloc tree deletion
after merge_reloc_roots"), reloc tree cleanup is delayed until one block
group is balanced.
Since fresh reloc roots are recorded during merge, as long as there
is no power loss, those orphan reloc roots converted from fresh ones are
handled without problem.
However when power loss happens, orphan reloc roots can be recorded
on-disk, thus at next mount time, we will have orphan reloc roots from
on-disk data directly, and ignored by clean_dirty_subvols() routine.
Then when background balance starts to balance another block group, and
needs to create new reloc root for the same root, btrfs_insert_item()
returns -EEXIST, and trigger that BUG_ON().
[FIX]
For orphan reloc roots, also queue them to rc->dirty_subvol_roots, so
all reloc roots no matter orphan or not, can be cleaned up properly and
avoid above BUG_ON().
And to cooperate with above change, clean_dirty_subvols() will check if
the queued root is a reloc root or a subvol root.
For a subvol root, do the old work, and for a orphan reloc root, clean it
up.
Fixes: d2311e6985 ("btrfs: relocation: Delay reloc tree deletion after merge_reloc_roots")
CC: stable@vger.kernel.org # 5.1
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When doing an incremental send we can now issue clone operations with a
source range that ends at the source's file eof and with a destination
range that ends at an offset smaller then the destination's file eof.
If the eof of the source file is not aligned to the sector size of the
filesystem, the receiver will get a -EINVAL error when trying to do the
operation or, on older kernels, silently corrupt the destination file.
The corruption happens on kernels without commit ac765f83f1
("Btrfs: fix data corruption due to cloning of eof block"), while the
failure to clone happens on kernels with that commit.
Example reproducer:
$ mkfs.btrfs -f /dev/sdb
$ mount /dev/sdb /mnt/sdb
$ xfs_io -f -c "pwrite -S 0xb1 0 2M" /mnt/sdb/foo
$ xfs_io -f -c "pwrite -S 0xc7 0 2M" /mnt/sdb/bar
$ xfs_io -f -c "pwrite -S 0x4d 0 2M" /mnt/sdb/baz
$ xfs_io -f -c "pwrite -S 0xe2 0 2M" /mnt/sdb/zoo
$ btrfs subvolume snapshot -r /mnt/sdb /mnt/sdb/base
$ btrfs send -f /tmp/base.send /mnt/sdb/base
$ xfs_io -c "reflink /mnt/sdb/bar 1560K 500K 100K" /mnt/sdb/bar
$ xfs_io -c "reflink /mnt/sdb/bar 1560K 0 100K" /mnt/sdb/zoo
$ xfs_io -c "truncate 550K" /mnt/sdb/bar
$ btrfs subvolume snapshot -r /mnt/sdb /mnt/sdb/incr
$ btrfs send -f /tmp/incr.send -p /mnt/sdb/base /mnt/sdb/incr
$ mkfs.btrfs -f /dev/sdc
$ mount /dev/sdc /mnt/sdc
$ btrfs receive -f /tmp/base.send /mnt/sdc
$ btrfs receive -vv -f /tmp/incr.send /mnt/sdc
(...)
truncate bar size=563200
utimes bar
clone zoo - source=bar source offset=512000 offset=0 length=51200
ERROR: failed to clone extents to zoo
Invalid argument
The failure happens because the clone source range ends at the eof of file
bar, 563200, which is not aligned to the filesystems sector size (4Kb in
this case), and the destination range ends at offset 0 + 51200, which is
less then the size of the file zoo (2Mb).
So fix this by detecting such case and instead of issuing a clone
operation for the whole range, do a clone operation for smaller range
that is sector size aligned followed by a write operation for the block
containing the eof. Here we will always be pessimistic and assume the
destination filesystem of the send stream has the largest possible sector
size (64Kb), since we have no way of determining it.
This fixes a recent regression introduced in kernel 5.2-rc1.
Fixes: 040ee6120c ("Btrfs: send, improve clone range")
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When using the no-holes feature, if we have a file with prealloc extents
with a start offset beyond the file's eof, doing an incremental send can
cause corruption of the file due to incorrect hole detection. Such case
requires that the prealloc extent(s) exist in both the parent and send
snapshots, and that a hole is punched into the file that covers all its
extents that do not cross the eof boundary.
Example reproducer:
$ mkfs.btrfs -f -O no-holes /dev/sdb
$ mount /dev/sdb /mnt/sdb
$ xfs_io -f -c "pwrite -S 0xab 0 500K" /mnt/sdb/foobar
$ xfs_io -c "falloc -k 1200K 800K" /mnt/sdb/foobar
$ btrfs subvolume snapshot -r /mnt/sdb /mnt/sdb/base
$ btrfs send -f /tmp/base.snap /mnt/sdb/base
$ xfs_io -c "fpunch 0 500K" /mnt/sdb/foobar
$ btrfs subvolume snapshot -r /mnt/sdb /mnt/sdb/incr
$ btrfs send -p /mnt/sdb/base -f /tmp/incr.snap /mnt/sdb/incr
$ md5sum /mnt/sdb/incr/foobar
816df6f64deba63b029ca19d880ee10a /mnt/sdb/incr/foobar
$ mkfs.btrfs -f /dev/sdc
$ mount /dev/sdc /mnt/sdc
$ btrfs receive -f /tmp/base.snap /mnt/sdc
$ btrfs receive -f /tmp/incr.snap /mnt/sdc
$ md5sum /mnt/sdc/incr/foobar
cf2ef71f4a9e90c2f6013ba3b2257ed2 /mnt/sdc/incr/foobar
--> Different checksum, because the prealloc extent beyond the
file's eof confused the hole detection code and it assumed
a hole starting at offset 0 and ending at the offset of the
prealloc extent (1200Kb) instead of ending at the offset
500Kb (the file's size).
Fix this by ensuring we never cross the file's size when issuing the
write operations for a hole.
Fixes: 16e7549f04 ("Btrfs: incompatible format change to remove hole extents")
CC: stable@vger.kernel.org # 3.14+
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Recent FITRIM work, namely bbbf7243d6 ("btrfs: combine device update
operations during transaction commit") combined the way certain
operations are recoded in a transaction. As a result an ASSERT was added
in dev_replace_finish to ensure the new code works correctly.
Unfortunately I got reports that it's possible to trigger the assert,
meaning that during a device replace it's possible to have an unfinished
chunk allocation on the source device.
This is supposed to be prevented by the fact that a transaction is
committed before finishing the replace oepration and alter acquiring the
chunk mutex. This is not sufficient since by the time the transaction is
committed and the chunk mutex acquired it's possible to allocate a chunk
depending on the workload being executed on the replaced device. This
bug has been present ever since device replace was introduced but there
was never code which checks for it.
The correct way to fix is to ensure that there is no pending device
modification operation when the chunk mutex is acquire and if there is
repeat transaction commit. Unfortunately it's not possible to just
exclude the source device from btrfs_fs_devices::dev_alloc_list since
this causes ENOSPC to be hit in transaction commit.
Fixing that in another way would need to add special cases to handle the
last writes and forbid new ones. The looped transaction fix is more
obvious, and can be easily backported. The runtime of dev-replace is
long so there's no noticeable delay caused by that.
Reported-by: David Sterba <dsterba@suse.com>
Fixes: 391cd9df81 ("Btrfs: fix unprotected alloc list insertion during the finishing procedure of replace")
CC: stable@vger.kernel.org # 4.4+
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Proc filesystem has special locking rules for various files. Thus
fanotify which opens files on event delivery can easily deadlock
against another process that waits for fanotify permission event to be
handled. Since permission events on /proc have doubtful value anyway,
just disallow them.
Link: https://lore.kernel.org/linux-fsdevel/20190320131642.GE9485@quack2.suse.cz/
Reviewed-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
The fuse_writeback_range() helper flushes dirty data to the userspace
filesystem.
When the function returns, the WRITE requests for the data in the given
range have all been completed. This is not equivalent to fsync() on the
given range, since the userspace filesystem may not yet have the data on
stable storage.
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Prior to sending COPY_FILE_RANGE to userspace filesystem, we must flush all
dirty pages in both the source and destination files.
This patch adds the missing flush of the source file.
Tested on libfuse-3.5.0 with:
libfuse/example/passthrough_ll /mnt/fuse/ -o writeback
libfuse/test/test_syscalls /mnt/fuse/tmp/test
Fixes: 88bc7d5097 ("fuse: add support for copy_file_range()")
Cc: <stable@vger.kernel.org> # v4.20
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Since xattr entry names are sorted, we don't have
to continue when current entry name is greater than
target.
Signed-off-by: Chengguang Xu <cgxu519@zoho.com.cn>
Signed-off-by: Jan Kara <jack@suse.cz>
Introduce new helper ext2_xattr_cmp_entry() for xattr
entry comparison.
Signed-off-by: Chengguang Xu <cgxu519@zoho.com.cn>
Signed-off-by: Jan Kara <jack@suse.cz>
We have introduced ext2_xattr_entry_valid() for xattr
entry sanity check, so it's better to do relevant things
in one place.
Signed-off-by: Chengguang Xu <cgxu519@zoho.com.cn>
Signed-off-by: Jan Kara <jack@suse.cz>
Calling bdi_rw_congested() instead of calling
bdi_read_congested() and bdi_write_congested().
Signed-off-by: Chengguang Xu <cgxu519@zoho.com.cn>
Signed-off-by: Jan Kara <jack@suse.cz>
In the vfs_statx() context, during path lookup, the dentry gets
added to sd->s_dentry via configfs_attach_attr(). In the end,
vfs_statx() kills the dentry by calling path_put(), which invokes
configfs_d_iput(). Ideally, this dentry must be removed from
sd->s_dentry but it doesn't if the sd->s_count >= 3. As a result,
sd->s_dentry is holding reference to a stale dentry pointer whose
memory is already freed up. This results in use-after-free issue,
when this stale sd->s_dentry is accessed later in
configfs_readdir() path.
This issue can be easily reproduced, by running the LTP test case -
sh fs_racer_file_list.sh /config
(https://github.com/linux-test-project/ltp/blob/master/testcases/kernel/fs/racer/fs_racer_file_list.sh)
Fixes: 76ae281f63 ('configfs: fix race between dentry put and lookup')
Signed-off-by: Sahitya Tummala <stummala@codeaurora.org>
Signed-off-by: Christoph Hellwig <hch@lst.de>
The function force_sigsegv is always called on the current task
so passing in current is redundant and not passing in current
makes this fact obvious.
This also makes it clear force_sigsegv always calls force_sig
on the current task.
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
The locking in force_sig_info is not prepared to deal with a task that
exits or execs (as sighand may change). The is not a locking problem
in force_sig as force_sig is only built to handle synchronous
exceptions.
Further the function force_sig_info changes the signal state if the
signal is ignored, or blocked or if SIGNAL_UNKILLABLE will prevent the
delivery of the signal. The signal SIGKILL can not be ignored and can
not be blocked and SIGNAL_UNKILLABLE won't prevent it from being
delivered.
So using force_sig rather than send_sig for SIGKILL is confusing
and pointless.
Because it won't impact the sending of the signal and and because
using force_sig is wrong, replace force_sig with send_sig.
Cc: Namjae Jeon <namjae.jeon@samsung.com>
Cc: Jeff Layton <jlayton@primarydata.com>
Cc: Steve French <smfrench@gmail.com>
Fixes: a5c3e1c725 ("Revert "cifs: No need to send SIGKILL to demux_thread during umount"")
Fixes: e7ddee9037 ("cifs: disable sharing session and tcon and add new TCP sharing code")
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
When hidden gendisk is revalidated, there's no point in revalidating
associated block device as there's none. We would thus just create new
bdev inode, report "detected capacity change from 0 to XXX" message and
evict the bdev inode again. Avoid this pointless dance and confusing
message in the kernel log.
Reviewed-by: Hannes Reinecke <hare@suse.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
In the FOPEN_DIRECT_IO case the write path doesn't call file_remove_privs()
and that means setuid bit is not cleared if unpriviliged user writes to a
file with setuid bit set.
pjdfstest chmod test 12.t tests this and fails.
Fix this by adding a flag to the FUSE_WRITE message that requests clearing
privileges on the given file. This needs
This better than just calling fuse_remove_privs(), because the attributes
may not be up to date, so in that case a write may miss clearing the
privileges.
Test case:
$ passthrough_ll /mnt/pasthrough-mnt -o default_permissions,allow_other,cache=never
$ mkdir /mnt/pasthrough-mnt/testdir
$ cd /mnt/pasthrough-mnt/testdir
$ prove -rv pjdfstests/tests/chmod/12.t
Reported-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Tested-by: Vivek Goyal <vgoyal@redhat.com>
They are the extended version of FS_IOC_FS[SG]ETFLAGS ioctls.
xfs_io -c "chattr <flags>" uses the new ioctls for setting flags.
This used to work in kernel pre v4.19, before stacked file ops
introduced the ovl_ioctl whitelist.
Reported-by: Dave Chinner <david@fromorbit.com>
Fixes: d1d04ef857 ("ovl: stack file ops")
Cc: <stable@vger.kernel.org> # v4.19
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
If io_copy_iov() fails, it will break the loop and report success,
albeit partially completed operation.
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
