firefly-linux-kernel-4.4.55.git
10 years agoext4: move error report out of atomic context in ext4_init_block_bitmap()
Dmitry Monakhov [Mon, 13 Oct 2014 07:42:12 +0000 (03:42 -0400)]
ext4: move error report out of atomic context in ext4_init_block_bitmap()

Error report likely result in IO so it is bad idea to do it from
atomic context.

This patch should fix following issue:

BUG: sleeping function called from invalid context at include/linux/buffer_head.h:349
in_atomic(): 1, irqs_disabled(): 0, pid: 137, name: kworker/u128:1
5 locks held by kworker/u128:1/137:
 #0:  ("writeback"){......}, at: [<ffffffff81085618>] process_one_work+0x228/0x4d0
 #1:  ((&(&wb->dwork)->work)){......}, at: [<ffffffff81085618>] process_one_work+0x228/0x4d0
 #2:  (jbd2_handle){......}, at: [<ffffffff81242622>] start_this_handle+0x712/0x7b0
 #3:  (&ei->i_data_sem){......}, at: [<ffffffff811fa387>] ext4_map_blocks+0x297/0x430
 #4:  (&(&bgl->locks[i].lock)->rlock){......}, at: [<ffffffff811f3180>] ext4_read_block_bitmap_nowait+0x5d0/0x630
CPU: 3 PID: 137 Comm: kworker/u128:1 Not tainted 3.17.0-rc2-00184-g82752e4 #165
Hardware name: Intel Corporation W2600CR/W2600CR, BIOS SE5C600.86B.99.99.x028.061320111235 06/13/2011
Workqueue: writeback bdi_writeback_workfn (flush-1:0)
 0000000000000411 ffff880813777288 ffffffff815c7fdc ffff880813777288
 ffff880813a8bba0 ffff8808137772a8 ffffffff8108fb30 ffff880803e01e38
 ffff880803e01e38 ffff8808137772c8 ffffffff811a8d53 ffff88080ecc6000
Call Trace:
 [<ffffffff815c7fdc>] dump_stack+0x51/0x6d
 [<ffffffff8108fb30>] __might_sleep+0xf0/0x100
 [<ffffffff811a8d53>] __sync_dirty_buffer+0x43/0xe0
 [<ffffffff811a8e03>] sync_dirty_buffer+0x13/0x20
 [<ffffffff8120f581>] ext4_commit_super+0x1d1/0x230
 [<ffffffff8120fa03>] save_error_info+0x23/0x30
 [<ffffffff8120fd06>] __ext4_error+0xb6/0xd0
 [<ffffffff8120f260>] ? ext4_group_desc_csum+0x140/0x190
 [<ffffffff811f2d8c>] ext4_read_block_bitmap_nowait+0x1dc/0x630
 [<ffffffff8122e23a>] ext4_mb_init_cache+0x21a/0x8f0
 [<ffffffff8113ae95>] ? lru_cache_add+0x55/0x60
 [<ffffffff8112e16c>] ? add_to_page_cache_lru+0x6c/0x80
 [<ffffffff8122eaa0>] ext4_mb_init_group+0x190/0x280
 [<ffffffff8122ec51>] ext4_mb_good_group+0xc1/0x190
 [<ffffffff8123309a>] ext4_mb_regular_allocator+0x17a/0x410
 [<ffffffff8122c821>] ? ext4_mb_use_preallocated+0x31/0x380
 [<ffffffff81233535>] ? ext4_mb_new_blocks+0x205/0x8e0
 [<ffffffff8116ed5c>] ? kmem_cache_alloc+0xfc/0x180
 [<ffffffff812335b0>] ext4_mb_new_blocks+0x280/0x8e0
 [<ffffffff8116f2c4>] ? __kmalloc+0x144/0x1c0
 [<ffffffff81221797>] ? ext4_find_extent+0x97/0x320
 [<ffffffff812257f4>] ext4_ext_map_blocks+0xbc4/0x1050
 [<ffffffff811fa387>] ? ext4_map_blocks+0x297/0x430
 [<ffffffff811fa3ab>] ext4_map_blocks+0x2bb/0x430
 [<ffffffff81200e43>] ? ext4_init_io_end+0x23/0x50
 [<ffffffff811feb44>] ext4_writepages+0x564/0xaf0
 [<ffffffff815cde3b>] ? _raw_spin_unlock+0x2b/0x40
 [<ffffffff810ac7bd>] ? lock_release_non_nested+0x2fd/0x3c0
 [<ffffffff811a009e>] ? writeback_sb_inodes+0x10e/0x490
 [<ffffffff811a009e>] ? writeback_sb_inodes+0x10e/0x490
 [<ffffffff811377e3>] do_writepages+0x23/0x40
 [<ffffffff8119c8ce>] __writeback_single_inode+0x9e/0x280
 [<ffffffff811a026b>] writeback_sb_inodes+0x2db/0x490
 [<ffffffff811a0664>] wb_writeback+0x174/0x2d0
 [<ffffffff810ac359>] ? lock_release_holdtime+0x29/0x190
 [<ffffffff811a0863>] wb_do_writeback+0xa3/0x200
 [<ffffffff811a0a40>] bdi_writeback_workfn+0x80/0x230
 [<ffffffff81085618>] ? process_one_work+0x228/0x4d0
 [<ffffffff810856cd>] process_one_work+0x2dd/0x4d0
 [<ffffffff81085618>] ? process_one_work+0x228/0x4d0
 [<ffffffff81085c1d>] worker_thread+0x35d/0x460
 [<ffffffff810858c0>] ? process_one_work+0x4d0/0x4d0
 [<ffffffff810858c0>] ? process_one_work+0x4d0/0x4d0
 [<ffffffff8108a885>] kthread+0xf5/0x100
 [<ffffffff810990e5>] ? local_clock+0x25/0x30
 [<ffffffff8108a790>] ? __init_kthread_worker+0x70/0x70
 [<ffffffff815ce2ac>] ret_from_fork+0x7c/0xb0
 [<ffffffff8108a790>] ? __init_kthread_work

Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
10 years agoext4: Replace open coded mdata csum feature to helper function
Dmitry Monakhov [Mon, 13 Oct 2014 07:36:16 +0000 (03:36 -0400)]
ext4: Replace open coded mdata csum feature to helper function

Besides the fact that this replacement improves code readability
it also protects from errors caused direct EXT4_S(sb)->s_es manipulation
which may result attempt to use uninitialized  csum machinery.

#Testcase_BEGIN
IMG=/dev/ram0
MNT=/mnt
mkfs.ext4 $IMG
mount $IMG $MNT
#Enable feature directly on disk, on mounted fs
tune2fs -O metadata_csum  $IMG
# Provoke metadata update, likey result in OOPS
touch $MNT/test
umount $MNT
#Testcase_END

# Replacement script
@@
expression E;
@@
- EXT4_HAS_RO_COMPAT_FEATURE(E, EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)
+ ext4_has_metadata_csum(E)

https://bugzilla.kernel.org/show_bug.cgi?id=82201

Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
10 years agoext4: delete useless comments about ext4_move_extents
Xiaoguang Wang [Sat, 11 Oct 2014 23:56:34 +0000 (19:56 -0400)]
ext4: delete useless comments about ext4_move_extents

In patch 'ext4: refactor ext4_move_extents code base',  Dmitry Monakhov has
refactored ext4_move_extents' implementation, but forgot to update the
corresponding comments, this patch will try to delete some useless comments.

Reviewed-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Xiaoguang Wang <wangxg.fnst@cn.fujitsu.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: fix reservation overflow in ext4_da_write_begin
Eric Sandeen [Sat, 11 Oct 2014 23:51:17 +0000 (19:51 -0400)]
ext4: fix reservation overflow in ext4_da_write_begin

Delalloc write journal reservations only reserve 1 credit,
to update the inode if necessary.  However, it may happen
once in a filesystem's lifetime that a file will cross
the 2G threshold, and require the LARGE_FILE feature to
be set in the superblock as well, if it was not set already.

This overruns the transaction reservation, and can be
demonstrated simply on any ext4 filesystem without the LARGE_FILE
feature already set:

dd if=/dev/zero of=testfile bs=1 seek=2147483646 count=1 \
conv=notrunc of=testfile
sync
dd if=/dev/zero of=testfile bs=1 seek=2147483647 count=1 \
conv=notrunc of=testfile

leads to:

EXT4-fs: ext4_do_update_inode:4296: aborting transaction: error 28 in __ext4_handle_dirty_super
EXT4-fs error (device loop0) in ext4_do_update_inode:4301: error 28
EXT4-fs error (device loop0) in ext4_reserve_inode_write:4757: Readonly filesystem
EXT4-fs error (device loop0) in ext4_dirty_inode:4876: error 28
EXT4-fs error (device loop0) in ext4_da_write_end:2685: error 28

Adjust the number of credits based on whether the flag is
already set, and whether the current write may extend past the
LARGE_FILE limit.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Cc: stable@vger.kernel.org
10 years agoext4: add ext4_iget_normal() which is to be used for dir tree lookups
Theodore Ts'o [Mon, 6 Oct 2014 02:56:00 +0000 (22:56 -0400)]
ext4: add ext4_iget_normal() which is to be used for dir tree lookups

If there is a corrupted file system which has directory entries that
point at reserved, metadata inodes, prohibit them from being used by
treating them the same way we treat Boot Loader inodes --- that is,
mark them to be bad inodes.  This prohibits them from being opened,
deleted, or modified via chmod, chown, utimes, etc.

In particular, this prevents a corrupted file system which has a
directory entry which points at the journal inode from being deleted
and its blocks released, after which point Much Hilarity Ensues.

Reported-by: Sami Liedes <sami.liedes@iki.fi>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
10 years agoext4: don't orphan or truncate the boot loader inode
Theodore Ts'o [Mon, 6 Oct 2014 02:47:07 +0000 (22:47 -0400)]
ext4: don't orphan or truncate the boot loader inode

The boot loader inode (inode #5) should never be visible in the
directory hierarchy, but it's possible if the file system is corrupted
that there will be a directory entry that points at inode #5.  In
order to avoid accidentally trashing it, when such a directory inode
is opened, the inode will be marked as a bad inode, so that it's not
possible to modify (or read) the inode from userspace.

Unfortunately, when we unlink this (invalid/illegal) directory entry,
we will put the bad inode on the ophan list, and then when try to
unlink the directory, we don't actually remove the bad inode from the
orphan list before freeing in-memory inode structure.  This means the
in-memory orphan list is corrupted, leading to a kernel oops.

In addition, avoid truncating a bad inode in ext4_destroy_inode(),
since truncating the boot loader inode is not a smart thing to do.

Reported-by: Sami Liedes <sami.liedes@iki.fi>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
10 years agoext4: grab missed write_count for EXT4_IOC_SWAP_BOOT
Dmitry Monakhov [Fri, 3 Oct 2014 16:47:23 +0000 (12:47 -0400)]
ext4: grab missed write_count for EXT4_IOC_SWAP_BOOT

Otherwise this provokes complain like follows:
WARNING: CPU: 12 PID: 5795 at fs/ext4/ext4_jbd2.c:48 ext4_journal_check_start+0x4e/0xa0()
Modules linked in: brd iTCO_wdt lpc_ich mfd_core igb ptp dm_mirror dm_region_hash dm_log dm_mod
CPU: 12 PID: 5795 Comm: python Not tainted 3.17.0-rc2-00175-gae5344f #158
Hardware name: Intel Corporation W2600CR/W2600CR, BIOS SE5C600.86B.99.99.x028.061320111235 06/13/2011
 0000000000000030 ffff8808116cfd28 ffffffff815c7dfc 0000000000000030
 0000000000000000 ffff8808116cfd68 ffffffff8106ce8c ffff8808116cfdc8
 ffff880813b16000 ffff880806ad6ae8 ffffffff81202008 0000000000000000
Call Trace:
 [<ffffffff815c7dfc>] dump_stack+0x51/0x6d
 [<ffffffff8106ce8c>] warn_slowpath_common+0x8c/0xc0
 [<ffffffff81202008>] ? ext4_ioctl+0x9e8/0xeb0
 [<ffffffff8106ceda>] warn_slowpath_null+0x1a/0x20
 [<ffffffff8122867e>] ext4_journal_check_start+0x4e/0xa0
 [<ffffffff81228c10>] __ext4_journal_start_sb+0x90/0x110
 [<ffffffff81202008>] ext4_ioctl+0x9e8/0xeb0
 [<ffffffff8107b0bd>] ? ptrace_stop+0x24d/0x2f0
 [<ffffffff81088530>] ? alloc_pid+0x480/0x480
 [<ffffffff8107b1f2>] ? ptrace_do_notify+0x92/0xb0
 [<ffffffff81186545>] do_vfs_ioctl+0x4e5/0x550
 [<ffffffff815cdbcb>] ? _raw_spin_unlock_irq+0x2b/0x40
 [<ffffffff81186603>] SyS_ioctl+0x53/0x80
 [<ffffffff815ce2ce>] tracesys+0xd0/0xd5

Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
10 years agoext4: optimize block allocation on grow indepth
Dmitry Monakhov [Thu, 2 Oct 2014 02:57:09 +0000 (22:57 -0400)]
ext4: optimize block allocation on grow indepth

It is reasonable to prepend newly created index to older one.

[ Dropped no longer used function parameter newext. -tytso ]

Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: get rid of code duplication
Dmitry Monakhov [Thu, 2 Oct 2014 02:26:17 +0000 (22:26 -0400)]
ext4: get rid of code duplication

Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: fix over-defensive complaint after journal abort
Dmitry Monakhov [Thu, 2 Oct 2014 02:23:15 +0000 (22:23 -0400)]
ext4: fix over-defensive complaint after journal abort

Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: fix return value of ext4_do_update_inode
Li Xi [Thu, 2 Oct 2014 02:11:06 +0000 (22:11 -0400)]
ext4: fix return value of ext4_do_update_inode

When ext4_do_update_inode() gets error from ext4_inode_blocks_set(),
error number should be returned.

Signed-off-by: Li Xi <lixi@ddn.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Reviewed-by: Jan Kara <jack@suse.cz>
10 years agoext4: fix mmap data corruption when blocksize < pagesize
Jan Kara [Thu, 2 Oct 2014 01:49:46 +0000 (21:49 -0400)]
ext4: fix mmap data corruption when blocksize < pagesize

Use truncate_isize_extended() when hole is being created in a file so that
->page_mkwrite() will get called for the partial tail page if it is
mmaped (see the first patch in the series for details).

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
10 years agovfs: fix data corruption when blocksize < pagesize for mmaped data
Jan Kara [Thu, 2 Oct 2014 01:49:18 +0000 (21:49 -0400)]
vfs: fix data corruption when blocksize < pagesize for mmaped data

->page_mkwrite() is used by filesystems to allocate blocks under a page
which is becoming writeably mmapped in some process' address space. This
allows a filesystem to return a page fault if there is not enough space
available, user exceeds quota or similar problem happens, rather than
silently discarding data later when writepage is called.

However VFS fails to call ->page_mkwrite() in all the cases where
filesystems need it when blocksize < pagesize. For example when
blocksize = 1024, pagesize = 4096 the following is problematic:
  ftruncate(fd, 0);
  pwrite(fd, buf, 1024, 0);
  map = mmap(NULL, 1024, PROT_WRITE, MAP_SHARED, fd, 0);
  map[0] = 'a';       ----> page_mkwrite() for index 0 is called
  ftruncate(fd, 10000); /* or even pwrite(fd, buf, 1, 10000) */
  mremap(map, 1024, 10000, 0);
  map[4095] = 'a';    ----> no page_mkwrite() called

At the moment ->page_mkwrite() is called, filesystem can allocate only
one block for the page because i_size == 1024. Otherwise it would create
blocks beyond i_size which is generally undesirable. But later at
->writepage() time, we also need to store data at offset 4095 but we
don't have block allocated for it.

This patch introduces a helper function filesystems can use to have
->page_mkwrite() called at all the necessary moments.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
10 years agoext4: fold ext4_nojournal_sops into ext4_sops
Theodore Ts'o [Thu, 18 Sep 2014 21:12:30 +0000 (17:12 -0400)]
ext4: fold ext4_nojournal_sops into ext4_sops

There's no longer any need to have a separate set of super_operations
for nojournal mode.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: support freezing ext2 (nojournal) file systems
Theodore Ts'o [Thu, 18 Sep 2014 21:12:02 +0000 (17:12 -0400)]
ext4: support freezing ext2 (nojournal) file systems

Through an oversight, when we added nojournal support to ext4, we
didn't add support to allow file system freezing.  This is relatively
easy to add, so let's do it.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reported-by: Dexuan Cui <decui@microsoft.com>
10 years agoext4: fold ext4_sync_fs_nojournal() into ext4_sync_fs()
Theodore Ts'o [Thu, 18 Sep 2014 20:12:37 +0000 (16:12 -0400)]
ext4: fold ext4_sync_fs_nojournal() into ext4_sync_fs()

This allows us to eliminate duplicate code, and eventually allow us to
also fold ext4_sops and ext4_nojournal_sops together.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: don't check quota format when there are no quota files
Jan Kara [Thu, 18 Sep 2014 05:12:15 +0000 (01:12 -0400)]
ext4: don't check quota format when there are no quota files

The check whether quota format is set even though there are no
quota files with journalled quota is pointless and it actually
makes it impossible to turn off journalled quotas (as there's
no way to unset journalled quota format). Just remove the check.

CC: stable@vger.kernel.org
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agojbd2: simplify calling convention around __jbd2_journal_clean_checkpoint_list
Jan Kara [Thu, 18 Sep 2014 04:58:12 +0000 (00:58 -0400)]
jbd2: simplify calling convention around __jbd2_journal_clean_checkpoint_list

__jbd2_journal_clean_checkpoint_list() returns number of buffers it
freed but noone was using the value so just stop doing that. This
also allows for simplifying the calling convention for
journal_clean_once_cp_list().

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agojbd2: avoid pointless scanning of checkpoint lists
Jan Kara [Thu, 18 Sep 2014 04:42:16 +0000 (00:42 -0400)]
jbd2: avoid pointless scanning of checkpoint lists

Yuanhan has reported that when he is running fsync(2) heavy workload
creating new files over ramdisk, significant amount of time is spent in
__jbd2_journal_clean_checkpoint_list() trying to clean old transactions
(but they cannot be cleaned up because flusher hasn't yet checkpointed
those buffers). The workload can be generated by:
  fs_mark -d /fs/ram0/1 -D 2 -N 2560 -n 1000000 -L 1 -S 1 -s 4096

Reduce the amount of scanning by stopping to scan the transaction list
once we find a transaction that cannot be checkpointed. Note that this
way of cleaning is still enough to keep freeing space in the journal
after fully checkpointed transactions.

Reported-and-tested-by: Yuanhan Liu <yuanhan.liu@linux.intel.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: explicitly inform user about orphan list cleanup
Dmitry Monakhov [Tue, 16 Sep 2014 18:52:03 +0000 (14:52 -0400)]
ext4: explicitly inform user about orphan list cleanup

Production fs likely compiled/mounted w/o jbd debugging, so orphan
list clearing will be silent.

Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agojbd2: jbd2_log_wait_for_space improve error detetcion
Dmitry Monakhov [Tue, 16 Sep 2014 18:50:50 +0000 (14:50 -0400)]
jbd2: jbd2_log_wait_for_space improve error detetcion

If EIO happens after we have dropped j_state_lock, we won't notice
that the journal has been aborted.  So it is reasonable to move this
check after we have grabbed the j_checkpoint_mutex and re-grabbed the
j_state_lock.  This patch helps to prevent false positive complain
after EIO.

#DMESG:
__jbd2_log_wait_for_space: needed 8448 blocks and only had 8386 space available
__jbd2_log_wait_for_space: no way to get more journal space in ram1-8
------------[ cut here ]------------
WARNING: CPU: 15 PID: 6739 at fs/jbd2/checkpoint.c:168 __jbd2_log_wait_for_space+0x188/0x200()
Modules linked in: brd iTCO_wdt lpc_ich mfd_core igb ptp dm_mirror dm_region_hash dm_log dm_mod
CPU: 15 PID: 6739 Comm: fsstress Tainted: G        W      3.17.0-rc2-00429-g684de57 #139
Hardware name: Intel Corporation W2600CR/W2600CR, BIOS SE5C600.86B.99.99.x028.061320111235 06/13/2011
 00000000000000a8 ffff88077aaab878 ffffffff815c1a8c 00000000000000a8
 0000000000000000 ffff88077aaab8b8 ffffffff8106ce8c ffff88077aaab898
 ffff8807c57e6000 ffff8807c57e6028 0000000000002100 ffff8807c57e62f0
Call Trace:
 [<ffffffff815c1a8c>] dump_stack+0x51/0x6d
 [<ffffffff8106ce8c>] warn_slowpath_common+0x8c/0xc0
 [<ffffffff8106ceda>] warn_slowpath_null+0x1a/0x20
 [<ffffffff812419f8>] __jbd2_log_wait_for_space+0x188/0x200
 [<ffffffff8123be9a>] start_this_handle+0x4da/0x7b0
 [<ffffffff810990e5>] ? local_clock+0x25/0x30
 [<ffffffff810aba87>] ? lockdep_init_map+0xe7/0x180
 [<ffffffff8123c5bc>] jbd2__journal_start+0xdc/0x1d0
 [<ffffffff811f2414>] ? __ext4_new_inode+0x7f4/0x1330
 [<ffffffff81222a38>] __ext4_journal_start_sb+0xf8/0x110
 [<ffffffff811f2414>] __ext4_new_inode+0x7f4/0x1330
 [<ffffffff810ac359>] ? lock_release_holdtime+0x29/0x190
 [<ffffffff812025bb>] ext4_create+0x8b/0x150
 [<ffffffff8117fe3b>] vfs_create+0x7b/0xb0
 [<ffffffff8118097b>] do_last+0x7db/0xcf0
 [<ffffffff8117e31d>] ? inode_permission+0x4d/0x50
 [<ffffffff811845d2>] path_openat+0x242/0x590
 [<ffffffff81191a76>] ? __alloc_fd+0x36/0x140
 [<ffffffff81184a6a>] do_filp_open+0x4a/0xb0
 [<ffffffff81191b61>] ? __alloc_fd+0x121/0x140
 [<ffffffff81172f20>] do_sys_open+0x170/0x220
 [<ffffffff8117300e>] SyS_open+0x1e/0x20
 [<ffffffff811715d6>] SyS_creat+0x16/0x20
 [<ffffffff815c7e12>] system_call_fastpath+0x16/0x1b
---[ end trace cd71c831f82059db ]---

Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agojbd2: free bh when descriptor block checksum fails
Darrick J. Wong [Tue, 16 Sep 2014 18:43:09 +0000 (14:43 -0400)]
jbd2: free bh when descriptor block checksum fails

Free the buffer head if the journal descriptor block fails checksum
verification.

This is the jbd2 port of the e2fsprogs patch "e2fsck: free bh on csum
verify error in do_one_pass".

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Cc: stable@vger.kernel.org
10 years agoext4: check EA value offset when loading
Darrick J. Wong [Tue, 16 Sep 2014 18:34:59 +0000 (14:34 -0400)]
ext4: check EA value offset when loading

When loading extended attributes, check each entry's value offset to
make sure it doesn't collide with the entries.

Without this check it is easy to crash the kernel by mounting a
malicious FS containing a file with an EA wherein e_value_offs = 0 and
e_value_size > 0 and then deleting the EA, which corrupts the name
list.

(See the f_ea_value_crash test's FS image in e2fsprogs for an example.)

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
10 years agoext4: don't keep using page if inline conversion fails
Darrick J. Wong [Thu, 11 Sep 2014 15:45:12 +0000 (11:45 -0400)]
ext4: don't keep using page if inline conversion fails

If inline->extent conversion fails (most probably due to ENOSPC) and
we release the temporary page that we allocated to transfer the file
contents, don't keep using the page pointer after releasing the page.
This occasionally leads to complaints about evicting locked pages or
hangs when blocksize > pagesize, because it's possible for the page to
get reallocated elsewhere in the meantime.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Tao Ma <tm@tao.ma>
10 years agoext4: validate external journal superblock checksum
Darrick J. Wong [Thu, 11 Sep 2014 15:44:36 +0000 (11:44 -0400)]
ext4: validate external journal superblock checksum

If the external journal device has metadata_csum enabled, verify
that the superblock checksum matches the block before we try to
mount.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agojbd2: fix journal checksum feature flag handling
Darrick J. Wong [Thu, 11 Sep 2014 15:38:21 +0000 (11:38 -0400)]
jbd2: fix journal checksum feature flag handling

Clear all three journal checksum feature flags before turning on
whichever journal checksum options we want.  Rearrange the error
checking so that newer flags get complained about first.

Reported-by: TR Reardon <thomas_reardon@hotmail.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: provide separate operations for sysfs feature files
Lukas Czerner [Thu, 11 Sep 2014 15:27:58 +0000 (11:27 -0400)]
ext4: provide separate operations for sysfs feature files

Currently sysfs feature files uses ext4_attr_ops as the file operations
to show/store data. However the feature files is not supposed to contain
any data at all, the sole existence of the file means that the module
support the feature. Moreover, none of the sysfs feature attributes
actually register show/store functions so that would not be a problem.

However if a sysfs feature attribute register a show or store function
we might be in trouble because the kobject in this case is _not_ embedded
in the ext4_sb_info structure as ext4_attr_show/store expect.

So just to be safe, provide separate empty sysfs_ops to use in
ext4_feat_ktype. This might safe us from potential problems in the
future. As a bonus we can "store" something more descriptive than
nothing in the files, so let it contain "enabled" to make it clear that
the feature is really present in the module.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: add sysfs entry showing whether the fs contains errors
Lukas Czerner [Thu, 11 Sep 2014 15:18:13 +0000 (11:18 -0400)]
ext4: add sysfs entry showing whether the fs contains errors

Currently there is no easy way to tell that the mounted file system
contains errors other than checking for log messages, or reading the
information directly from superblock.

This patch adds new sysfs entries:

errors_count (number of fs errors we encounter)
first_error_time (unix timestamp for the first error we see)
last_error_time (unix timestamp for the last error we see)

If the file system is not marked as containing errors then any of the
file will return 0. Otherwise it will contain valid information. More
details about the errors should as always be found in the logs.

Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: don't use MAXQUOTAS value
Jan Kara [Thu, 11 Sep 2014 15:15:15 +0000 (11:15 -0400)]
ext4: don't use MAXQUOTAS value

MAXQUOTAS value defines maximum number of quota types VFS supports.
This isn't necessarily the number of types ext4 supports. Although
ext4 will support project quotas, use ext4 private definition for
consistency with other filesystems.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agojbd/jbd2: use non-movable memory for the jbd superblock
Gioh Kim [Fri, 5 Sep 2014 02:36:35 +0000 (22:36 -0400)]
jbd/jbd2: use non-movable memory for the jbd superblock

Sicne the jbd/jbd2 superblock is not released until the file system is
unmounted, allocate the buffer cache from the non-moveable area to
allow page migration and CMA allocations to more easily succeed.

Signed-off-by: Gioh Kim <gioh.kim@lge.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jan Kara <jack@suse.cz>
10 years agoext4: use non-movable memory for the ext4 superblock
Gioh Kim [Fri, 5 Sep 2014 02:36:15 +0000 (22:36 -0400)]
ext4: use non-movable memory for the ext4 superblock

Since the ext4 superblock is not released until the file system is
unmounted, allocate the buffer cache entry for the ext4 superblock out
of the non-moveable are to allow page migrations and thus CMA
allocations to more easily succeed if the CMA area is limited.

Signed-off-by: Gioh Kim <gioh.kim@lge.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jan Kara <jack@suse.cz>
10 years agofs/buffer.c: support buffer cache allocations with gfp modifiers
Gioh Kim [Fri, 5 Sep 2014 02:04:42 +0000 (22:04 -0400)]
fs/buffer.c: support buffer cache allocations with gfp modifiers

A buffer cache is allocated from movable area because it is referred
for a while and released soon.  But some filesystems are taking buffer
cache for a long time and it can disturb page migration.

New APIs are introduced to allocate buffer cache with user specific
flag.  *_gfp APIs are for user want to set page allocation flag for
page cache allocation.  And *_unmovable APIs are for the user wants to
allocate page cache from non-movable area.

Signed-off-by: Gioh Kim <gioh.kim@lge.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jan Kara <jack@suse.cz>
10 years agoext4: renumber EXT4_EX_* flags to avoid flag aliasing problems
Theodore Ts'o [Thu, 4 Sep 2014 22:09:29 +0000 (18:09 -0400)]
ext4: renumber EXT4_EX_* flags to avoid flag aliasing problems

Suggested-by: Andreas Dilger <adilger@dilger.ca>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agojbd2: optimize jbd2_log_do_checkpoint() a bit
Jan Kara [Thu, 4 Sep 2014 22:09:29 +0000 (18:09 -0400)]
jbd2: optimize jbd2_log_do_checkpoint() a bit

When we discover written out buffer in transaction checkpoint list we
don't have to recheck validity of a transaction. Either this is the
last buffer in a transaction - and then we are done - or this isn't
and then we can just take another buffer from the checkpoint list
without dropping j_list_lock.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agojbd2: don't call get_bh() before calling __jbd2_journal_remove_checkpoint()
Theodore Ts'o [Thu, 4 Sep 2014 22:09:22 +0000 (18:09 -0400)]
jbd2: don't call get_bh() before calling __jbd2_journal_remove_checkpoint()

The __jbd2_journal_remove_checkpoint() doesn't require an elevated
b_count; indeed, until the jh structure gets released by the call to
jbd2_journal_put_journal_head(), the bh's b_count is elevated by
virtue of the existence of the jh structure.

Suggested-by: Jan Kara <jack@suse.cz>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: drop the EXT4_STATE_DELALLOC_RESERVED flag
Theodore Ts'o [Thu, 4 Sep 2014 22:08:22 +0000 (18:08 -0400)]
ext4: drop the EXT4_STATE_DELALLOC_RESERVED flag

Having done a full regression test, we can now drop the
DELALLOC_RESERVED state flag.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jan Kara <jack@suse.cz>
10 years agoext4: prepare to drop EXT4_STATE_DELALLOC_RESERVED
Theodore Ts'o [Thu, 4 Sep 2014 22:07:25 +0000 (18:07 -0400)]
ext4: prepare to drop EXT4_STATE_DELALLOC_RESERVED

The EXT4_STATE_DELALLOC_RESERVED flag was originally implemented
because it was too hard to make sure the mballoc and get_block flags
could be reliably passed down through all of the codepaths that end up
calling ext4_mb_new_blocks().

Since then, we have mb_flags passed down through most of the code
paths, so getting rid of EXT4_STATE_DELALLOC_RESERVED isn't as tricky
as it used to.

This commit plumbs in the last of what is required, and then adds a
WARN_ON check to make sure we haven't missed anything.  If this passes
a full regression test run, we can then drop
EXT4_STATE_DELALLOC_RESERVED.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jan Kara <jack@suse.cz>
10 years agoext4: pass allocation_request struct to ext4_(alloc,splice)_branch
Theodore Ts'o [Thu, 4 Sep 2014 22:06:25 +0000 (18:06 -0400)]
ext4: pass allocation_request struct to ext4_(alloc,splice)_branch

Instead of initializing the allocation_request structure in
ext4_alloc_branch(), set it up in ext4_ind_map_blocks(), and then pass
it to ext4_alloc_branch() and ext4_splice_branch().

This allows ext4_ind_map_blocks to pass flags in the allocation
request structure without having to add Yet Another argument to
ext4_alloc_branch().

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jan Kara <jack@suse.cz>
10 years agoext4: track extent status tree shrinker delay statictics
Zheng Liu [Tue, 2 Sep 2014 02:26:49 +0000 (22:26 -0400)]
ext4: track extent status tree shrinker delay statictics

This commit adds some statictics in extent status tree shrinker.  The
purpose to add these is that we want to collect more details when we
encounter a stall caused by extent status tree shrinker.  Here we count
the following statictics:
  stats:
    the number of all objects on all extent status trees
    the number of reclaimable objects on lru list
    cache hits/misses
    the last sorted interval
    the number of inodes on lru list
  average:
    scan time for shrinking some objects
    the number of shrunk objects
  maximum:
    the inode that has max nr. of objects on lru list
    the maximum scan time for shrinking some objects

The output looks like below:
  $ cat /proc/fs/ext4/sda1/es_shrinker_info
  stats:
    28228 objects
    6341 reclaimable objects
    5281/631 cache hits/misses
    586 ms last sorted interval
    250 inodes on lru list
  average:
    153 us scan time
    128 shrunk objects
  maximum:
    255 inode (255 objects, 198 reclaimable)
    125723 us max scan time

If the lru list has never been sorted, the following line will not be
printed:
    586ms last sorted interval
If there is an empty lru list, the following lines also will not be
printed:
    250 inodes on lru list
  ...
  maximum:
    255 inode (255 objects, 198 reclaimable)
    0 us max scan time

Meanwhile in this commit a new trace point is defined to print some
details in __ext4_es_shrink().

Cc: Andreas Dilger <adilger.kernel@dilger.ca>
Cc: Jan Kara <jack@suse.cz>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: improve extents status tree trace point
Zheng Liu [Tue, 2 Sep 2014 02:22:13 +0000 (22:22 -0400)]
ext4: improve extents status tree trace point

This commit improves the trace point of extents status tree.  We rename
trace_ext4_es_shrink_enter in ext4_es_count() because it is also used
in ext4_es_scan() and we can not identify them from the result.

Further this commit fixes a variable name in trace point in order to
keep consistency with others.

Cc: Andreas Dilger <adilger.kernel@dilger.ca>
Cc: Jan Kara <jack@suse.cz>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Zheng Liu <wenqing.lz@taobao.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: fix comments about get_blocks
Seunghun Lee [Tue, 2 Sep 2014 02:15:30 +0000 (22:15 -0400)]
ext4: fix comments about get_blocks

get_blocks is renamed to get_block.

Signed-off-by: Seunghun Lee <waydi1@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: enable block_validity by default
Darrick J. Wong [Tue, 2 Sep 2014 01:34:09 +0000 (21:34 -0400)]
ext4: enable block_validity by default

Enable by default the block_validity feature, which checks for
collisions between newly allocated blocks and critical system
metadata.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agojbd2: fold __wait_cp_io into jbd2_log_do_checkpoint()
Theodore Ts'o [Tue, 2 Sep 2014 01:26:09 +0000 (21:26 -0400)]
jbd2: fold __wait_cp_io into jbd2_log_do_checkpoint()

__wait_cp_io() is only called by jbd2_log_do_checkpoint().  Fold it in
to make it a bit easier to understand.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agojbd2: fold __process_buffer() into jbd2_log_do_checkpoint()
Theodore Ts'o [Tue, 2 Sep 2014 01:19:01 +0000 (21:19 -0400)]
jbd2: fold __process_buffer() into jbd2_log_do_checkpoint()

__process_buffer() is only called by jbd2_log_do_checkpoint(), and it
had a very complex locking protocol where it would be called with the
j_list_lock, and sometimes exit with the lock held (if the return code
was 0), or release the lock.

This was confusing both to humans and to smatch (which erronously
complained that the lock was taken twice).

Folding __process_buffer() to the caller allows us to simplify the
control flow, making the resulting function easier to read and reason
about, and dropping the compiled size of fs/jbd2/checkpoint.c by 150
bytes (over 4% of the text size).

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Reviewed-by: Jan Kara <jack@suse.cz>
10 years agoext4: rename ext4_ext_find_extent() to ext4_find_extent()
Theodore Ts'o [Mon, 1 Sep 2014 18:43:09 +0000 (14:43 -0400)]
ext4: rename ext4_ext_find_extent() to ext4_find_extent()

Make the function name less redundant.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: reuse path object in ext4_move_extents()
Theodore Ts'o [Mon, 1 Sep 2014 18:42:09 +0000 (14:42 -0400)]
ext4: reuse path object in ext4_move_extents()

Reuse the path object in ext4_move_extents() so we don't unnecessarily
free and reallocate it.

Also clean up the get_ext_path() wrapper so that it has the same
semantics of freeing the path object on error as ext4_ext_find_extent().

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: reuse path object in ext4_ext_shift_extents()
Theodore Ts'o [Mon, 1 Sep 2014 18:41:09 +0000 (14:41 -0400)]
ext4: reuse path object in ext4_ext_shift_extents()

Now that the semantics of ext4_ext_find_extent() are much cleaner,
it's safe and more efficient to reuse the path object across the
multiple calls to ext4_ext_find_extent() in ext4_ext_shift_extents().

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: teach ext4_ext_find_extent() to realloc path if necessary
Theodore Ts'o [Mon, 1 Sep 2014 18:40:09 +0000 (14:40 -0400)]
ext4: teach ext4_ext_find_extent() to realloc path if necessary

This adds additional safety in case for some reason we end reusing a
path structure which isn't big enough for current depth of the inode.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: allow a NULL argument to ext4_ext_drop_refs()
Theodore Ts'o [Mon, 1 Sep 2014 18:39:09 +0000 (14:39 -0400)]
ext4: allow a NULL argument to ext4_ext_drop_refs()

Teach ext4_ext_drop_refs() to accept a NULL argument, much like
kfree().  This allows us to drop a lot of checks to make sure path is
non-NULL before calling ext4_ext_drop_refs().

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: call ext4_ext_drop_refs() from ext4_ext_find_extent()
Theodore Ts'o [Mon, 1 Sep 2014 18:38:09 +0000 (14:38 -0400)]
ext4: call ext4_ext_drop_refs() from ext4_ext_find_extent()

In nearly all of the calls to ext4_ext_find_extent() where the caller
is trying to recycle the path object, ext4_ext_drop_refs() gets called
to release the buffer heads before the path object gets overwritten.
To simplify things for the callers, and to avoid the possibility of a
memory leak, make ext4_ext_find_extent() responsible for dropping the
buffers.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: drop EXT4_EX_NOFREE_ON_ERR from rest of extents handling code
Theodore Ts'o [Mon, 1 Sep 2014 18:37:09 +0000 (14:37 -0400)]
ext4: drop EXT4_EX_NOFREE_ON_ERR from rest of extents handling code

Drop EXT4_EX_NOFREE_ON_ERR from ext4_ext_create_new_leaf(),
ext4_split_extent(), ext4_convert_unwritten_extents_endio().

This requires fixing all of their callers to potentially
ext4_ext_find_extent() to free the struct ext4_ext_path object in case
of an error, and there are interlocking dependencies all the way up to
ext4_ext_map_blocks(), ext4_swap_extents(), and
ext4_ext_remove_space().

Once this is done, we can drop the EXT4_EX_NOFREE_ON_ERR flag since it
is no longer necessary.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: drop EXT4_EX_NOFREE_ON_ERR in convert_initialized_extent()
Theodore Ts'o [Mon, 1 Sep 2014 18:36:09 +0000 (14:36 -0400)]
ext4: drop EXT4_EX_NOFREE_ON_ERR in convert_initialized_extent()

Transfer responsibility of freeing struct ext4_ext_path on error to
ext4_ext_find_extent().

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: collapse ext4_convert_initialized_extents()
Theodore Ts'o [Mon, 1 Sep 2014 18:35:09 +0000 (14:35 -0400)]
ext4: collapse ext4_convert_initialized_extents()

The function ext4_convert_initialized_extents() is only called by a
single function --- ext4_ext_convert_initalized_extents().  Inline the
code and get rid of the unnecessary bits in order to simplify the code.

Rename ext4_ext_convert_initalized_extents() to
convert_initalized_extents() since it's a static function that is
actually only used in a single caller, ext4_ext_map_blocks().

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: teach ext4_ext_find_extent() to free path on error
Theodore Ts'o [Mon, 1 Sep 2014 18:34:09 +0000 (14:34 -0400)]
ext4: teach ext4_ext_find_extent() to free path on error

Right now, there are a places where it is all to easy to leak memory
on an error path, via a usage like this:

struct ext4_ext_path *path = NULL

while (...) {
...
path = ext4_ext_find_extent(inode, block, path, 0);
if (IS_ERR(path)) {
/* oops, if path was non-NULL before the call to
   ext4_ext_find_extent, we've leaked it!  :-(  */
...
return PTR_ERR(path);
}
...
}

Unfortunately, there some code paths where we are doing the following
instead:

path = ext4_ext_find_extent(inode, block, orig_path, 0);

and where it's important that we _not_ free orig_path in the case
where ext4_ext_find_extent() returns an error.

So change the function signature of ext4_ext_find_extent() so that it
takes a struct ext4_ext_path ** for its third argument, and by
default, on an error, it will free the struct ext4_ext_path, and then
zero out the struct ext4_ext_path * pointer.  In order to avoid
causing problems, we add a flag EXT4_EX_NOFREE_ON_ERR which causes
ext4_ext_find_extent() to use the original behavior of forcing the
caller to deal with freeing the original path pointer on the error
case.

The goal is to get rid of EXT4_EX_NOFREE_ON_ERR entirely, but this
allows for a gentle transition and makes the patches easier to verify.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: fix accidental flag aliasing in ext4_map_blocks flags
Theodore Ts'o [Mon, 1 Sep 2014 18:33:09 +0000 (14:33 -0400)]
ext4: fix accidental flag aliasing in ext4_map_blocks flags

Commit b8a8684502a0f introduced an accidental flag aliasing between
EXT4_EX_NOCACHE and EXT4_GET_BLOCKS_CONVERT_UNWRITTEN.

Fortunately, this didn't introduce any untorward side effects --- we
got lucky.  Nevertheless, fix this and leave a warning to hopefully
avoid this from happening in the future.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: fix ZERO_RANGE bug hidden by flag aliasing
Theodore Ts'o [Mon, 1 Sep 2014 18:32:09 +0000 (14:32 -0400)]
ext4: fix ZERO_RANGE bug hidden by flag aliasing

We accidently aliased EXT4_EX_NOCACHE and EXT4_GET_CONVERT_UNWRITTEN
falgs, which apparently was hiding a bug that was unmasked when this
flag aliasing issue was addressed (see the subsequent commit).  The
reproduction case was:

   fsx -N 10000 -l 500000 -r 4096 -t 4096 -w 4096 -Z -R -W /vdb/junk

... which would cause fsx to report corruption in the data file.

The fix we have is a bit of an overkill, but I'd much rather be
conservative for now, and we can optimize ZERO_RANGE_FL handling
later.  The fact that we need to zap the extent_status cache for the
inode is unfortunate, but correctness is far more important than
performance.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: Namjae Jeon <namjae.jeon@samsung.com>
10 years agoext4: fix ext4_swap_extents() error handling
Theodore Ts'o [Sun, 31 Aug 2014 19:03:14 +0000 (15:03 -0400)]
ext4: fix ext4_swap_extents() error handling

If ext4_ext_find_extent() returns an error, we have to clear path1 or
path2 or else we would end up trying to free an ERR_PTR, which would
be bad.

Also eliminate some redundant code and mark the error paths as unlikely()

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: refactor ext4_move_extents code base
Dmitry Monakhov [Sun, 31 Aug 2014 03:52:19 +0000 (23:52 -0400)]
ext4: refactor ext4_move_extents code base

ext4_move_extents is too complex for review. It has duplicate almost
each function available in the rest of other codebase. It has useless
artificial restriction orig_offset == donor_offset. But in fact logic
of ext4_move_extents is very simple:

Iterate extents one by one (similar to ext4_fill_fiemap_extents)
   ->Iterate each page covered extent (similar to generic_perform_write)
     ->swap extents for covered by page (can be shared with IOC_MOVE_DATA)

Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: use ext4_ext_next_allocated_block instead of mext_next_extent
Dmitry Monakhov [Sun, 31 Aug 2014 03:50:56 +0000 (23:50 -0400)]
ext4: use ext4_ext_next_allocated_block instead of mext_next_extent

This allows us to make mext_next_extent static and potentially get rid
of it.

Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: use ext4_update_i_disksize instead of opencoded ones
Dmitry Monakhov [Sun, 31 Aug 2014 03:34:06 +0000 (23:34 -0400)]
ext4: use ext4_update_i_disksize instead of opencoded ones

Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: remove a duplicate call in ext4_init_new_dir()
Wang Shilong [Sat, 30 Aug 2014 03:20:44 +0000 (23:20 -0400)]
ext4: remove a duplicate call in ext4_init_new_dir()

ext4_journal_get_write_access() has just been called in ext4_append()
calling it again here is duplicated.

Signed-off-by: Wang Shilong <wshilong@ddn.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: convert do_split() to use the ERR_PTR convention
Theodore Ts'o [Sat, 30 Aug 2014 00:52:18 +0000 (20:52 -0400)]
ext4: convert do_split() to use the ERR_PTR convention

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: convert dx_probe() to use the ERR_PTR convention
Theodore Ts'o [Sat, 30 Aug 2014 00:52:17 +0000 (20:52 -0400)]
ext4: convert dx_probe() to use the ERR_PTR convention

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: convert ext4_bread() to use the ERR_PTR convention
Theodore Ts'o [Sat, 30 Aug 2014 00:52:15 +0000 (20:52 -0400)]
ext4: convert ext4_bread() to use the ERR_PTR convention

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: convert ext4_getblk() to use the ERR_PTR convention
Theodore Ts'o [Sat, 30 Aug 2014 00:51:32 +0000 (20:51 -0400)]
ext4: convert ext4_getblk() to use the ERR_PTR convention

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoext4: convert ext4_dx_find_entry() to use the ERR_PTR convention
Theodore Ts'o [Sat, 30 Aug 2014 00:49:51 +0000 (20:49 -0400)]
ext4: convert ext4_dx_find_entry() to use the ERR_PTR convention

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
10 years agoMerge tag 'ext4_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Fri, 29 Aug 2014 18:52:46 +0000 (11:52 -0700)]
Merge tag 'ext4_for_linus_stable' of git://git./linux/kernel/git/tytso/ext4

Pull ext4 bugfixes from Ted Ts'o:
 "Ext4 bug fixes for 3.17, to provide better handling of memory
  allocation failures, and to fix some journaling bugs involving
  journal checksums and FALLOC_FL_ZERO_RANGE"

* tag 'ext4_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
  ext4: fix same-dir rename when inline data directory overflows
  jbd2: fix descriptor block size handling errors with journal_csum
  jbd2: fix infinite loop when recovering corrupt journal blocks
  ext4: update i_disksize coherently with block allocation on error path
  ext4: fix transaction issues for ext4_fallocate and ext_zero_range
  ext4: fix incorect journal credits reservation in ext4_zero_range
  ext4: move i_size,i_disksize update routines to helper function
  ext4: fix BUG_ON in mb_free_blocks()
  ext4: propagate errors up to ext4_find_entry()'s callers

10 years agoMerge tag 'dm-3.17-fix' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mappe...
Linus Torvalds [Fri, 29 Aug 2014 18:49:10 +0000 (11:49 -0700)]
Merge tag 'dm-3.17-fix' of git://git./linux/kernel/git/device-mapper/linux-dm

Pull device mapper fix from Mike Snitzer:
 "Fix a 3.17-rc1 regression introduced by switching the DM crypt target
  to using per-bio data"

* tag 'dm-3.17-fix' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm crypt: fix access beyond the end of allocated space

10 years agoMerge branch 'for-linus' of git://git.kernel.dk/linux-block
Linus Torvalds [Fri, 29 Aug 2014 18:21:48 +0000 (11:21 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-block

Pull block layer fixes from Jens Axboe:
 "A smaller collection of fixes that have come up since the initial
  merge window pull request.  This contains:

   - error handling cleanup and support for larger than 16 byte cdbs in
     sg_io() from Christoph.  The latter just matches what bsg and
     friends support, sg_io() got left out in the merge.

   - an option for brd to expose partitions in /proc/partitions.  They
     are hidden by default for compat reasons.  From Dmitry Monakhov.

   - a few blk-mq fixes from me - killing a dead/unused flag, fix for
     merging happening even if turned off, and correction of a few
     comments.

   - removal of unnecessary ->owner setting in systemace.  From Michal
     Simek.

   - two related fixes for a problem with nesting freezing of queues in
     blk-mq.  One from Ming Lei removing an unecessary freeze operation,
     and another from Tejun fixing the nesting regression introduced in
     the merge window.

   - fix for a BUG_ON() at bio_endio time when protection info is
     attached and the IO has an error.  From Sagi Grimberg.

   - two scsi_ioctl bug fixes for regressions with scsi-mq from Tony
     Battersby.

   - a cfq weight update fix and subsequent comment update from Toshiaki
     Makita"

* 'for-linus' of git://git.kernel.dk/linux-block:
  cfq-iosched: Add comments on update timing of weight
  cfq-iosched: Fix wrong children_weight calculation
  block: fix error handling in sg_io
  fix regression in SCSI_IOCTL_SEND_COMMAND
  scsi-mq: fix requests that use a separate CDB buffer
  block: support > 16 byte CDBs for SG_IO
  block: cleanup error handling in sg_io
  brd: add ram disk visibility option
  block: systemace: Remove .owner field for driver
  blk-mq: blk_mq_freeze_queue() should allow nesting
  blk-mq: correct a few wrong/bad comments
  block: Fix BUG_ON when pi errors occur
  blk-mq: don't allow merges if turned off for the queue
  blk-mq: get rid of unused BLK_MQ_F_SHOULD_SORT flag
  blk-mq: fix WARNING "percpu_ref_kill() called more than once!"

10 years agoalpha: io: implement relaxed accessor macros for writes
Will Deacon [Fri, 25 Jul 2014 00:53:54 +0000 (17:53 -0700)]
alpha: io: implement relaxed accessor macros for writes

write{b,w,l,q}_relaxed are implemented by some architectures in order to
permit memory-mapped I/O writes with weaker barrier semantics than the
non-relaxed variants.

This patch implements these write macros for Alpha, in the same vein as
the relaxed read macros, which are already implemented.

Acked-by: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoalpha: Wire up sched_setattr, sched_getattr, and renameat2 syscalls.
Michael Cree [Fri, 25 Jul 2014 00:53:53 +0000 (17:53 -0700)]
alpha: Wire up sched_setattr, sched_getattr, and renameat2 syscalls.

Signed-off-by: Michael Cree <mcree@orcon.net.nz>
Signed-off-by: Matt Turner <mattst88@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoext4: fix same-dir rename when inline data directory overflows
Darrick J. Wong [Wed, 27 Aug 2014 22:40:09 +0000 (18:40 -0400)]
ext4: fix same-dir rename when inline data directory overflows

When performing a same-directory rename, it's possible that adding or
setting the new directory entry will cause the directory to overflow
the inline data area, which causes the directory to be converted to an
extent-based directory.  Under this circumstance it is necessary to
re-read the directory when deleting the old dirent because the "old
directory" context still points to i_block in the inode table, which
is now an extent tree root!  The delete fails with an FS error, and
the subsequent fsck complains about incorrect link counts and
hardlinked directories.

Test case (originally found with flat_dir_test in the metadata_csum
test program):

# mkfs.ext4 -O inline_data /dev/sda
# mount /dev/sda /mnt
# mkdir /mnt/x
# touch /mnt/x/changelog.gz /mnt/x/copyright /mnt/x/README.Debian
# sync
# for i in /mnt/x/*; do mv $i $i.longer; done
# ls -la /mnt/x/
total 0
-rw-r--r-- 1 root root 0 Aug 25 12:03 changelog.gz.longer
-rw-r--r-- 1 root root 0 Aug 25 12:03 copyright
-rw-r--r-- 1 root root 0 Aug 25 12:03 copyright.longer
-rw-r--r-- 1 root root 0 Aug 25 12:03 README.Debian.longer

(Hey!  Why are there four files now??)

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
10 years agojbd2: fix descriptor block size handling errors with journal_csum
Darrick J. Wong [Wed, 27 Aug 2014 22:40:07 +0000 (18:40 -0400)]
jbd2: fix descriptor block size handling errors with journal_csum

It turns out that there are some serious problems with the on-disk
format of journal checksum v2.  The foremost is that the function to
calculate descriptor tag size returns sizes that are too big.  This
causes alignment issues on some architectures and is compounded by the
fact that some parts of jbd2 use the structure size (incorrectly) to
determine the presence of a 64bit journal instead of checking the
feature flags.

Therefore, introduce journal checksum v3, which enlarges the
descriptor block tag format to allow for full 32-bit checksums of
journal blocks, fix the journal tag function to return the correct
sizes, and fix the jbd2 recovery code to use feature flags to
determine 64bitness.

Add a few function helpers so we don't have to open-code quite so
many pieces.

Switching to a 16-byte block size was found to increase journal size
overhead by a maximum of 0.1%, to convert a 32-bit journal with no
checksumming to a 32-bit journal with checksum v3 enabled.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Reported-by: TR Reardon <thomas_reardon@hotmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
10 years agojbd2: fix infinite loop when recovering corrupt journal blocks
Darrick J. Wong [Wed, 27 Aug 2014 22:40:05 +0000 (18:40 -0400)]
jbd2: fix infinite loop when recovering corrupt journal blocks

When recovering the journal, don't fall into an infinite loop if we
encounter a corrupt journal block.  Instead, just skip the block and
return an error, which fails the mount and thus forces the user to run
a full filesystem fsck.

Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
10 years agoext4: update i_disksize coherently with block allocation on error path
Dmitry Monakhov [Wed, 27 Aug 2014 22:40:03 +0000 (18:40 -0400)]
ext4: update i_disksize coherently with block allocation on error path

In case of delalloc block i_disksize may be less than i_size. So we
have to update i_disksize each time we allocated and submitted some
blocks beyond i_disksize.  We weren't doing this on the error paths,
so fix this.

testcase: xfstest generic/019

Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
10 years agodm crypt: fix access beyond the end of allocated space
Mikulas Patocka [Thu, 28 Aug 2014 15:09:31 +0000 (11:09 -0400)]
dm crypt: fix access beyond the end of allocated space

The DM crypt target accesses memory beyond allocated space resulting in
a crash on 32 bit x86 systems.

This bug is very old (it dates back to 2.6.25 commit 3a7f6c990ad04 "dm
crypt: use async crypto").  However, this bug was masked by the fact
that kmalloc rounds the size up to the next power of two.  This bug
wasn't exposed until 3.17-rc1 commit 298a9fa08a ("dm crypt: use per-bio
data").  By switching to using per-bio data there was no longer any
padding beyond the end of a dm-crypt allocated memory block.

To minimize allocation overhead dm-crypt puts several structures into one
block allocated with kmalloc.  The block holds struct ablkcipher_request,
cipher-specific scratch pad (crypto_ablkcipher_reqsize(any_tfm(cc))),
struct dm_crypt_request and an initialization vector.

The variable dmreq_start is set to offset of struct dm_crypt_request
within this memory block.  dm-crypt allocates the block with this size:
cc->dmreq_start + sizeof(struct dm_crypt_request) + cc->iv_size.

When accessing the initialization vector, dm-crypt uses the function
iv_of_dmreq, which performs this calculation: ALIGN((unsigned long)(dmreq
+ 1), crypto_ablkcipher_alignmask(any_tfm(cc)) + 1).

dm-crypt allocated "cc->iv_size" bytes beyond the end of dm_crypt_request
structure.  However, when dm-crypt accesses the initialization vector, it
takes a pointer to the end of dm_crypt_request, aligns it, and then uses
it as the initialization vector.  If the end of dm_crypt_request is not
aligned on a crypto_ablkcipher_alignmask(any_tfm(cc)) boundary the
alignment causes the initialization vector to point beyond the allocated
space.

Fix this bug by calculating the variable iv_size_padding and adding it
to the allocated size.

Also correct the alignment of dm_crypt_request.  struct dm_crypt_request
is specific to dm-crypt (it isn't used by the crypto subsystem at all),
so it is aligned on __alignof__(struct dm_crypt_request).

Also align per_bio_data_size on ARCH_KMALLOC_MINALIGN, so that it is
aligned as if the block was allocated with kmalloc.

Reported-by: Krzysztof Kolasa <kkolasa@winsoft.pl>
Tested-by: Milan Broz <gmazyland@gmail.com>
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
10 years agoMerge tag 'backlight-fixes-3.17' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Thu, 28 Aug 2014 17:47:10 +0000 (10:47 -0700)]
Merge tag 'backlight-fixes-3.17' of git://git./linux/kernel/git/lee/backlight

Pull backlight fix from Lee Jones:
 "One simple fix to invalidate GPIO non-request"

* tag 'backlight-fixes-3.17' of git://git.kernel.org/pub/scm/linux/kernel/git/lee/backlight:
  pwm-backlight: Fix bogus request for GPIO#0 when instantiated from DT

10 years agoMerge tag 'mfd-fixes-3.17' of git://git.kernel.org/pub/scm/linux/kernel/git/lee/mfd
Linus Torvalds [Thu, 28 Aug 2014 17:46:25 +0000 (10:46 -0700)]
Merge tag 'mfd-fixes-3.17' of git://git./linux/kernel/git/lee/mfd

Pull mfd fixes from Lee Jones:
 "Couple of simple fixes due for the 3.17 rcs

  (and a sneaky document addition that slipped from the previous
  pull-request)"

* tag 'mfd-fixes-3.17' of git://git.kernel.org/pub/scm/linux/kernel/git/lee/mfd:
  mfd: twl4030-power: Fix PM idle pin configuration to not conflict with regulators
  mfd: tc3589x: Add device tree bindings
  mfd: ab8500-core: Use 'ifdef' for config options
  mfd: htc-i2cpld: Fix %d confusingly prefixed with 0x in format string
  mfd: omap-usb-host: Fix %d confusingly prefixed with 0x in format string

10 years agoMerge tag 'pinctrl-v3.17-2' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw...
Linus Torvalds [Thu, 28 Aug 2014 17:31:29 +0000 (10:31 -0700)]
Merge tag 'pinctrl-v3.17-2' of git://git./linux/kernel/git/linusw/linux-pinctrl

Pull pin-control fixes from Linus Walleij:
 "My first (a bit delayed) pack of pin control fixes for the v3.17
  series, only driver fixes:

   - SH-PFC (Renesas) r8a7791 CAN bus pin group problem
   - Rockchip (GPIO0 configuration)
   - Tegra-xusb (interrupt handling)
   - Exynos (GPIO interrupt locking)
   - Qualcomm (fix misleading example interrupts)
   - minor non-critical fixes for abx500 and AT91 also sneaked in,
     because I initially intended this pull for post RC-1, hope it's
     still OK"

* tag 'pinctrl-v3.17-2' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl:
  pinctrl: qcom: apq8064: Correct interrupts in example
  pinctrl: exynos: Lock GPIOs as interrupts when used as EINTs
  pinctrl: pinctrl-at91.c: fix decimal printf format specifiers prefixed with 0x
  pinctrl: abx500: remove useless check
  pinctrl: tegra-xusb: testing wrong variable in probe()
  pinctrl: tegra-xusb: fix an off by one test
  pinctrl: rockchip: fix rk3288 gpio0 configuration
  sh-pfc: r8a7791: fix CAN pin groups

10 years agoMerge tag 'for-3.17-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/sumits...
Linus Torvalds [Thu, 28 Aug 2014 17:30:25 +0000 (10:30 -0700)]
Merge tag 'for-3.17-rc3' of git://git./linux/kernel/git/sumits/dma-buf

Pull dma-buf fixes from Sumit Semwal:
 "The major changes for 3.17 already went via Greg-KH's tree this time
  as well; this is a small pull request for dma-buf - all documentation
  related"

* tag 'for-3.17-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/sumits/dma-buf:
  dma-buf/fence: Fix one more kerneldoc warning
  dma-buf/fence: Fix a kerneldoc warning
  Documentation/dma-buf-sharing.txt: update API descriptions

10 years agoMerge tag 'sound-3.17-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Thu, 28 Aug 2014 16:44:25 +0000 (09:44 -0700)]
Merge tag 'sound-3.17-rc3' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "Here contains not many exciting changes but just a few minor ones: An
  off-by-one proc write fix, a couple of trivial incldue guard fixes,
  Acer laptop pinconfig fix, and a fix for DSD formats that are still
  rarely used"

* tag 'sound-3.17-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: hda - Set up initial pins for Acer Aspire V5
  ALSA: pcm: Fix the silence data for DSD formats
  ALSA: ctxfi: ct20k1reg: Fix typo in include guard
  ALSA: hda: ca0132_regs.h: Fix typo in include guard
  ALSA: core: fix buffer overflow in snd_info_get_line()

10 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Thu, 28 Aug 2014 16:40:37 +0000 (09:40 -0700)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "Nothing major, one core oops fixes, some radeon oops fixes, some sti
  driver fixups, msm driver fixes and a minor Kconfig update for the ww
  mutex debugging"

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  drm/ast: Add missing entry to dclk_table[]
  drm: fix division-by-zero on dumb_create()
  ww-mutex: clarify help text for DEBUG_WW_MUTEX_SLOWPATH
  radeon: Test for PCI root bus before assuming bus->self
  drm/radeon: handle broken disabled rb mask gracefully (6xx/7xx) (v2)
  drm/radeon: save/restore the PD addr on suspend/resume
  drm/msm: Fix missing unlock on error in msm_fbdev_create()
  drm/msm: fix compile error for non-dt builds
  drm/msm/mdp4: request vblank during modeset
  drm/msm: avoid flood of kernel logs on faults
  drm: sti: Add missing dependency on RESET_CONTROLLER
  drm: sti: Make of_device_id array const
  drm: sti: Fix return value check in sti_drm_platform_probe()
  drm: sti: hda: fix return value check in sti_hda_probe()
  drm: sti: hdmi: fix return value check in sti_hdmi_probe()
  drm: sti: tvout: fix return value check in sti_tvout_probe()

10 years agomfd: twl4030-power: Fix PM idle pin configuration to not conflict with regulators
Tony Lindgren [Tue, 19 Aug 2014 15:24:05 +0000 (08:24 -0700)]
mfd: twl4030-power: Fix PM idle pin configuration to not conflict with regulators

Commit 43fef47f94a1 (mfd: twl4030-power: Add a configuration to turn
off oscillator during off-idle) added support for configuring the PMIC
to cut off resources during deeper idle states to save power.

This however caused regression for n900 display power that needed the
PMIC configuration to be disabled with commit d937678ab625 (ARM: dts:
Revert enabling of twl configuration for n900).

Turns out the root cause of the problem is that we must use
TWL4030_RESCONFIG_UNDEF instead of DEV_GRP_NULL to avoid disabling
regulators that may have been enabled before the init function
for twl4030-power.c runs. With TWL4030_RESCONFIG_UNDEF we let the
regulator framework control the regulators like it should. Here we
need to only configure the sys_clken and sys_off_mode triggers for
the regulators that cannot be done by the regulator framework as
it's not running at that point.

This allows us to enable the PMIC configuration for n900.

Fixes: 43fef47f94a1 (mfd: twl4030-power: Add a configuration to turn off oscillator during off-idle)
Cc: stable@vger.kernel.org # v3.16
Signed-off-by: Tony Lindgren <tony@atomide.com>
Tested-by: Aaro Koskinen <aaro.koskinen@iki.fi>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
10 years agomfd: tc3589x: Add device tree bindings
Linus Walleij [Wed, 9 Apr 2014 07:28:00 +0000 (09:28 +0200)]
mfd: tc3589x: Add device tree bindings

This defines the device tree bindings for the Toshiba TC3589x
series of multi-purpose expanders. Only the stuff I can test
is defined: GPIO and keypad. Others may implement more
subdevices further down the road.

This is to complement
commit a435ae1d51e2f18414f2a87219fdbe068231e692
"mfd: Enable the tc3589x for Device Tree" which left off
the definition of the device tree bindings.

Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
10 years agocfq-iosched: Add comments on update timing of weight
Toshiaki Makita [Thu, 28 Aug 2014 08:14:58 +0000 (17:14 +0900)]
cfq-iosched: Add comments on update timing of weight

Explain that weight has to be updated on activation.
This complements previous fix e15693ef18e1 ("cfq-iosched: Fix wrong
children_weight calculation").

Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Signed-off-by: Jens Axboe <axboe@fb.com>
10 years agodma-buf/fence: Fix one more kerneldoc warning
Thierry Reding [Fri, 8 Aug 2014 11:06:30 +0000 (13:06 +0200)]
dma-buf/fence: Fix one more kerneldoc warning

The seqno_fence_init() function's cond argument isn't described in the
kerneldoc comment. Fix that to silence a warning when building DocBook
documentation.

Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org>
10 years agodma-buf/fence: Fix a kerneldoc warning
Thierry Reding [Fri, 8 Aug 2014 10:42:32 +0000 (12:42 +0200)]
dma-buf/fence: Fix a kerneldoc warning

kerneldoc doesn't know how to parse variables, so don't let it try.

Signed-off-by: Thierry Reding <treding@nvidia.com>
Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org>
10 years agoDocumentation/dma-buf-sharing.txt: update API descriptions
Gioh Kim [Tue, 13 May 2014 23:49:43 +0000 (08:49 +0900)]
Documentation/dma-buf-sharing.txt: update API descriptions

Update some descriptions for API arguments and descriptions.

Signed-off-by: Gioh Kim <gioh.kim@lge.com>
Signed-off-by: Sumit Semwal <sumit.semwal@linaro.org>
10 years agodrm/ast: Add missing entry to dclk_table[]
Y.C. Chen [Fri, 22 Aug 2014 03:00:17 +0000 (11:00 +0800)]
drm/ast: Add missing entry to dclk_table[]

This avoid reading past the end of the list for certain modes

Signed-off-by: Y.C. Chen <yc_chen@aspeedtech.com>
Reviewed-by: Egbert Eich <eich@freedesktop.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agoMerge branch 'drm-3.17-rc2-sti-fixes' of git://git.linaro.org/people/benjamin.gaignar...
Dave Airlie [Thu, 28 Aug 2014 01:48:58 +0000 (11:48 +1000)]
Merge branch 'drm-3.17-rc2-sti-fixes' of git://git.linaro.org/people/benjamin.gaignard/kernel into drm-fixes

I have tested the 6 patches send on mailing list since you merge the sti driver.
I haven't seen issue with those patches except for the missing
dependency on Kconfig
where I have change "depends on" to "select".

* 'drm-3.17-rc2-sti-fixes' of git://git.linaro.org/people/benjamin.gaignard/kernel:
  drm: sti: Add missing dependency on RESET_CONTROLLER
  drm: sti: Make of_device_id array const
  drm: sti: Fix return value check in sti_drm_platform_probe()
  drm: sti: hda: fix return value check in sti_hda_probe()
  drm: sti: hdmi: fix return value check in sti_hdmi_probe()
  drm: sti: tvout: fix return value check in sti_tvout_probe()

10 years agoMerge branch 'msm-fixes-3.17' of git://people.freedesktop.org/~robclark/linux into...
Dave Airlie [Thu, 28 Aug 2014 01:48:05 +0000 (11:48 +1000)]
Merge branch 'msm-fixes-3.17' of git://people.freedesktop.org/~robclark/linux into drm-fixes

misc msm fixes from Rob.

* 'msm-fixes-3.17' of git://people.freedesktop.org/~robclark/linux:
  drm/msm: Fix missing unlock on error in msm_fbdev_create()
  drm/msm: fix compile error for non-dt builds
  drm/msm/mdp4: request vblank during modeset
  drm/msm: avoid flood of kernel logs on faults

10 years agodrm: fix division-by-zero on dumb_create()
David Herrmann [Sun, 24 Aug 2014 17:23:26 +0000 (19:23 +0200)]
drm: fix division-by-zero on dumb_create()

Kinda unexpected, but DIV_ROUND_UP() can overflow if passed an argument
bigger than UINT_MAX - DIVISOR. Fix this by testing for "!cpp" before
using it in the following division.

Note that DIV_ROUND_UP() is defined as:
        #define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d))

..this will obviously overflow if (n + d - 1) is bigger than UINT_MAX.

Reported-by: Tommi Rantala <tt.rantala@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Reviewed-by: Rob Clark <robdclark@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agoww-mutex: clarify help text for DEBUG_WW_MUTEX_SLOWPATH
Rob Clark [Wed, 27 Aug 2014 15:19:26 +0000 (11:19 -0400)]
ww-mutex: clarify help text for DEBUG_WW_MUTEX_SLOWPATH

We really don't want distro's enabling this in their kernels.  Try and
make that more clear.

Signed-off-by: Rob Clark <robdclark@gmail.com>
Acked-by: Maarten Lankhorst <maarten.lankhorst@canonical.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
10 years agoMerge branch 'drm-fixes-3.17' of git://people.freedesktop.org/~agd5f/linux into drm...
Dave Airlie [Thu, 28 Aug 2014 01:32:20 +0000 (11:32 +1000)]
Merge branch 'drm-fixes-3.17' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

Just a few more radeon fixes for 3.17.

* 'drm-fixes-3.17' of git://people.freedesktop.org/~agd5f/linux:
  radeon: Test for PCI root bus before assuming bus->self
  drm/radeon: handle broken disabled rb mask gracefully (6xx/7xx) (v2)
  drm/radeon: save/restore the PD addr on suspend/resume

10 years agoMerge branch 'sec-v3.17-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/sergeh...
Linus Torvalds [Thu, 28 Aug 2014 00:32:37 +0000 (17:32 -0700)]
Merge branch 'sec-v3.17-rc2' of git://git./linux/kernel/git/sergeh/linux-security

Pull tomoyo fix from Serge Hallyn.

* 'sec-v3.17-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/sergeh/linux-security:
  tomoyo: Fix pathname calculation breakage.

10 years agoext4: fix transaction issues for ext4_fallocate and ext_zero_range
Dmitry Monakhov [Wed, 27 Aug 2014 22:40:00 +0000 (18:40 -0400)]
ext4: fix transaction issues for ext4_fallocate and ext_zero_range

After commit f282ac19d86f we use different transactions for
preallocation and i_disksize update which result in complain from fsck
after power-failure.  spotted by generic/019. IMHO this is regression
because fs becomes inconsistent, even more 'e2fsck -p' will no longer
works (which drives admins go crazy) Same transaction requirement
applies ctime,mtime updates

testcase: xfstest generic/019

Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
10 years agoext4: fix incorect journal credits reservation in ext4_zero_range
Dmitry Monakhov [Wed, 27 Aug 2014 22:33:49 +0000 (18:33 -0400)]
ext4: fix incorect journal credits reservation in ext4_zero_range

Currently we reserve only 4 blocks but in worst case scenario
ext4_zero_partial_blocks() may want to zeroout and convert two
non adjacent blocks.

Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@vger.kernel.org
10 years agoradeon: Test for PCI root bus before assuming bus->self
Alex Williamson [Wed, 27 Aug 2014 19:01:35 +0000 (13:01 -0600)]
radeon: Test for PCI root bus before assuming bus->self

If we assign a Radeon device to a virtual machine, we can no longer
assume a fixed hardware topology, like the GPU having a parent device.
This patch simply adds a few pci_is_root_bus() tests to avoid passing
a NULL pointer to PCI access functions, allowing the radeon driver to
work in a QEMU 440FX machine with an assigned HD8570 on the emulated
PCI root bus.

Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
10 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid
Linus Torvalds [Wed, 27 Aug 2014 16:38:06 +0000 (09:38 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jikos/hid

Pull HID fixes from Jiri Kosina:

 - fixes for potential memory corruption problems in magicmouse and
   picolcd drivers (the HW would have to be manufactured to be
   deliberately evil to trigger those) which were found by Steven
   Vittitoe

 - fix for false error message appearing in dmesg from logitech-dj
   driver, from Benjamin Tissoires

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid:
  HID: picolcd: sanity check report size in raw_event() callback
  HID: magicmouse: sanity check report size in raw_event() callback
  HID: logitech-dj: prevent false errors to be shown

10 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux...
Linus Torvalds [Wed, 27 Aug 2014 16:14:17 +0000 (09:14 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/mason/linux-btrfs

Pull btrfs fixes from Chris Mason:
 "The biggest of these comes from Liu Bo, who tracked down a hang we've
  been hitting since moving to kernel workqueues (it's a btrfs bug, not
  in the generic code).  His patch needs backporting to 3.16 and 3.15
  stable, which I'll send once this is in.

  Otherwise these are assorted fixes.  Most were integrated last week
  during KS, but I wanted to give everyone the chance to test the
  result, so I waited for rc2 to come out before sending"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs: (24 commits)
  Btrfs: fix task hang under heavy compressed write
  Btrfs: fix filemap_flush call in btrfs_file_release
  Btrfs: fix crash on endio of reading corrupted block
  btrfs: fix leak in qgroup_subtree_accounting() error path
  btrfs: Use right extent length when inserting overlap extent map.
  Btrfs: clone, don't create invalid hole extent map
  Btrfs: don't monopolize a core when evicting inode
  Btrfs: fix hole detection during file fsync
  Btrfs: ensure tmpfile inode is always persisted with link count of 0
  Btrfs: race free update of commit root for ro snapshots
  Btrfs: fix regression of btrfs device replace
  Btrfs: don't consider the missing device when allocating new chunks
  Btrfs: Fix wrong device size when we are resizing the device
  Btrfs: don't write any data into a readonly device when scrub
  Btrfs: Fix the problem that the replace destroys the seed filesystem
  btrfs: Return right extent when fiemap gives unaligned offset and len.
  Btrfs: fix wrong extent mapping for DirectIO
  Btrfs: fix wrong write range for filemap_fdatawrite_range()
  Btrfs: fix wrong missing device counter decrease
  Btrfs: fix unzeroed members in fs_devices when creating a fs from seed fs
  ...