Currently ext4_ext_walk_space() only takes i_data_sem for read when
searching for the extent at given block with ext4_ext_find_extent().
Then it drops the lock and the extent tree can be changed at will.
However later on we're searching for the 'next' extent, but the extent
tree might already have changed, so the information might not be
accurate.

In fact we can hit BUG_ON(end <= start) if the extent got inserted into
the tree after the one we found and before the block we were searching
for. This has been reproduced by running xfstests 225 in loop on s390x
architecture, but theoretically we could hit this on any other
architecture as well, but probably not as often.

Moreover the extent currently in delayed allocation might be allocated
after we search the extent tree and before we search extent status tree
delayed buffers resulting in those delayed buffers being completely
missed, even though completely written and allocated.

We fix all those problems in several steps:

 1. remove unnecessary callback indirection
 2. rename functions
        ext4_ext_walk_space -> ext4_fill_fiemap_extents
        ext4_ext_fiemap_cb -> ext4_find_delayed_extent
 3. move fiemap_fill_next_extent() into ext4_fill_fiemap_extents()
 4. hold the i_data_sem for:
        ext4_ext_find_extent()
        ext4_ext_next_allocated_block()
        ext4_find_delayed_extent()
 5. call fiemap_fill_next_extent after releasing the i_data_sem
 6. move path reinitialization into the critical section.
Signed-off-by: NLukas Czerner <lczerner@redhat.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

The calls to ext4_jbd2_file_inode() are needed to guarantee that we do
not expose stale data in the data=ordered mode.  However, they are not
necessary because in all of the cases where we have newly allocated
blocks in the delayed allocation write path, we immediately submit the
dirty pages for I/O.  Hence, we can avoid the overhead of adding the
inode to the list of inodes whose data pages will be to be flushed out
to disk completely during the next commit operation.
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

ext4_da_block_invalidatepages is missing a pagevec_init(),
which means that pvec->cold contains random garbage.

This affects whether the page goes to the front or
back of the LRU when ->cold makes it to
free_hot_cold_page()
Reviewed-by: NLukas Czerner <lczerner@redhat.com>
Reviewed-by: NCarlos Maiolino <cmaiolino@redhat.com>
Signed-off-by: NEric Sandeen <sandeen@redhat.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>
Cc: stable@vger.kernel.org

During a directory entry lookup of a hashed directory, if the
hash-based lookup functions fail and we fall back to a linear scan,
don't try to verify the dirent checksum on the internal nodes of the
hash tree because they don't store a checksum in a hidden dirent like
the leaf nodes do.
Reported-by: NGeorge Spelvin <linux@horizon.com>
Signed-off-by: NDarrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

If there is no space for a checksum in a directory leaf node,
previously we would use EXT4_ERROR_INODE() which would mark the file
system as inconsistent.  While it would be nice to use e2fsck -D, it
certainly isn't required, so just print a warning using
ext4_warning().
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>
Cc: "Darrick J. Wong" <darrick.wong@oracle.com>

This patch makes ext4 really support SEEK_DATA/SEEK_HOLE flags.  Block-mapped
and extent-mapped files are fully implemented together because ext4_map_blocks
hides this differences.

After applying this patch, it will cause a failure in xfstest #285 when the file
is block-mapped due to block-mapped file isn't support fallocate(2).

I had tried to use ext4_ext_walk_space() to retrieve the offset for a
extent-mapped file.  But finally I decide to keep using ext4_map_blocks() to
support SEEK_DATA/SEEK_HOLE because ext4_map_blocks() can hide the difference
between block-mapped file and extent-mapped file.  Moreover, in next step,
extent status tree will track all extent status, and we can get all mappings
from this tree.  So I think that using ext4_map_blocks() is a better choice.

CC: Hugh Dickins <hughd@google.com>
Signed-off-by: NJie Liu <jeff.liu@oracle.com>
Signed-off-by: NZheng Liu <wenqing.lz@taobao.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

Signed-off-by: NYongqiang Yang <xiaoqiangnk@gmail.com>
Signed-off-by: NAllison Henderson <achender@linux.vnet.ibm.com>
Signed-off-by: NZheng Liu <wenqing.lz@taobao.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

Signed-off-by: NYongqiang Yang <xiaoqiangnk@gmail.com>
Signed-off-by: NAllison Henderson <achender@linux.vnet.ibm.com>
Signed-off-by: NZheng Liu <wenqing.lz@taobao.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

This patch adds some tracepoints in extent status tree.
Signed-off-by: NZheng Liu <wenqing.lz@taobao.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

This patch lets ext4 maintain extent status tree.

Currently it only tracks delay extent status in extent status tree.  When a
delay allocation is issued, the related delay extent will be inserted into
extent status tree.  When a delay extent is written out or invalidated, it will
be removed from this tree.
Signed-off-by: NYongqiang Yang <xiaoqiangnk@gmail.com>
Signed-off-by: NAllison Henderson <achender@linux.vnet.ibm.com>
Signed-off-by: NZheng Liu <wenqing.lz@taobao.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

Let ext4 initialize extent status tree of an inode.
Signed-off-by: NYongqiang Yang <xiaoqiangnk@gmail.com>
Signed-off-by: NAllison Henderson <achender@linux.vnet.ibm.com>
Signed-off-by: NZheng Liu <wenqing.lz@taobao.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

This patch adds operations on a extent status tree.

CC: Lukas Czerner <lczerner@redhat.com>
Signed-off-by: NYongqiang Yang <xiaoqiangnk@gmail.com>
Signed-off-by: NAllison Henderson <achender@linux.vnet.ibm.com>
Signed-off-by: NHugh Dickins <hughd@google.com>
Signed-off-by: NZheng Liu <wenqing.lz@taobao.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

This patch adds two structures that supports extent status tree, extent_status
and ext4_es_tree.  Currently extent_status is used to track a delay extent for
an inode, which record the start block and the length of the delay extent.
ext4_es_tree is used to store all extent_status for an inode in memory.
Signed-off-by: NYongqiang Yang <xiaoqiangnk@gmail.com>
Signed-off-by: NAllison Henderson <achender@linux.vnet.ibm.com>
Signed-off-by: NZheng Liu <wenqing.lz@taobao.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

There are some places in ext4_fill_super() where we would not return
proper error code if something fails. The confusion is caused probably
due to the fact that we have two "kind-of" return variables 'ret'and
'err'.

'ret' is used to return error code from ext4_fill_super() where err is
used to store return values from other functions within ext4_fill_super().
However some places were missing the obligatory 'ret = err'. We could
put the assignment where it is missing, but we can have better "future
proof" solution. Or we could convert the code to use just one, but it
would require more rewrites.

This commit fixes the problem by returning value from 'err' variable if
it is set and 'ret' otherwise in error handling branch of the
ext4_fill_super(). The reasoning is that 'ret' value is often set to
default "-EINVAL" or explicit value, where 'err' is used to store
return value from other functions and should be otherwise zero.

https://bugzilla.kernel.org/show_bug.cgi?id=48431Signed-off-by: NLukas Czerner <lczerner@redhat.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

In ext4_xattr_set_acl(), if ext4_journal_start() returns an error,
posix_acl_release() will not be called for 'acl' which may result in a
memory leak.

This patch fixes that.
Reviewed-by: NLukas Czerner <lczerner@redhat.com>
Signed-off-by: NEugene Shatokhin <eugene.shatokhin@rosalab.ru>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>
Cc: stable@vger.kernel.org

729f52c6 introduced function ext4_get_block_write_nolock() that
is very similar to _ext4_get_block(). Eliminate code duplication
by passing different flags to _ext4_get_block()

Tested: xfs tests
Reviewed-by: NZheng Liu <wenqing.lz@taobao.com>
Signed-off-by: NAnatol Pomozov <anatol.pomozov@gmail.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

Tested: xfs tests
Reviewed-by: NZheng Liu <wenqing.lz@taobao.com>
Signed-off-by: NAnatol Pomozov <anatol.pomozov@gmail.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

When ext4_ext_handle_uninitialized_extents(), we will directly return
from ext4_ext_map_blocks().  The trace point of
trace_ext4_ext_map_blocks_exit isn't called, and the user doesn't see
any result.  This patch tries to fix this problem.

Meanwhile in ext4_ext_handle_uninitialized_extents it returns errors
or the number of allocated blocks.  So 'ret' variable can be removed
due to previously modifications.
Signed-off-by: NZheng Liu <wenqing.lz@taobao.com>

When we use trace_ext4_ext/ind_map_blocks_exit, print the value of
map->m_flags in order that we can understand the extent's current
status.
Reviewed-by: NLukas Czerner <lczerner@redhat.com>
Signed-off-by: NZheng Liu <wenqing.lz@taobao.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

In trace_ext4_ext_handle_uninitialized_extents we don't care about the
value of map->m_flags because this value is probably 0, and we prefer
to get the value of flags because we can know how to handle this
extent in this function.
Reviewed-by: NLukas Czerner <lczerner@redhat.com>
Signed-off-by: NZheng Liu <wenqing.lz@taobao.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

We should warn user then the discard request fails. However we need to
exclude -EOPNOTSUPP case since parts of the device might not support it
while other parts can. So print the kernel warning when the error !=
-EOPNOTSUPP is returned from ext4_issue_discard().

We should also handle error cases in batched discard, again excluding
EOPNOTSUPP.
Reviewed-by: NCarlos Maiolino <cmaiolino@redhat.com>
Signed-off-by: NLukas Czerner <lczerner@redhat.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

Notify user when mounting the file system with -o discard option, but
the device does not support discard. Obviously we do not want to fail
the mount or disable the options, because the underlying device might
change in future even without file system remount.
Reviewed-by: NCarlos Maiolino <cmaiolino@redhat.com>
Signed-off-by: NLukas Czerner <lczerner@redhat.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

Signed-off-by: NAlan Cox <alan@linux.intel.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

Signed-off-by: NZhao Hongjiang <zhaohongjiang@huawei.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

The use of variable length arrays in structs (VLAIS) in the Linux Kernel code
precludes the use of compilers which don't implement VLAIS (for instance the
Clang compiler). Since ctx is always a 32-bit CRC, hard coding a size of 4
bytes accomplishes the same thing without the use of VLAIS. This is the same
technique already employed in fs/ext4/ext4.h
Signed-off-by: NMark Charlebois <charlebm@gmail.com>
Signed-off-by: NBehan Webster <behanw@converseincode.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

ext4_handle_release_buffer() was intended to remove journal
write access from a buffer, but it doesn't actually do anything
at all other than add a BUFFER_TRACE point, but it's not reliably
used for that either.  Remove all the associated dead code.
Signed-off-by: NEric Sandeen <sandeen@redhat.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>
Reviewed-by: NCarlos Maiolino <cmaiolino@redhat.com>

I think the whole function could be made prettier, but
that goto really took the cake for too-clever-by-half.
Signed-off-by: NEric Sandeen <sandeen@redhat.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

"overhead" was a write-only variable in this function after commit
952fc18e; we set it to 0 for minixdf, or to sbi->s_overhead if !minixdf,
but never read it again after that.

We need to use it, not sbi->s_overhead, when subtracting out overhead
for f_blocks, or we get the wrong answer for minixdf.
Signed-off-by: NEric Sandeen <sandeen@redhat.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

commit 119c0d44 changed
ext4_new_inode() such that the inode bitmap was being modified
outside a transaction, which could lead to corruption, and was
discovered when journal_checksum found a bad checksum in the
journal during log replay.

Nix ran into this when using the journal_async_commit mount
option, which enables journal checksumming.  The ensuing
journal replay failures due to the bad checksums led to
filesystem corruption reported as the now infamous
"Apparent serious progressive ext4 data corruption bug"

[ Changed by tytso to only call ext4_journal_get_write_access() only
  when we're fairly certain that we're going to allocate the inode. ]

I've tested this by mounting with journal_checksum and
running fsstress then dropping power; I've also tested by
hacking DM to create snapshots w/o first quiescing, which
allows me to test journal replay repeatedly w/o actually
power-cycling the box.  Without the patch I hit a journal
checksum error every time.  With this fix it survives
many iterations.
Reported-by: NNix <nix@esperi.org.uk>
Signed-off-by: NEric Sandeen <sandeen@redhat.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>
Cc: stable@vger.kernel.org

Pull ktest confusion fix from Steven Rostedt:
 "With the v3.7-rc2 kernel, the network cards on my target boxes were
  not being brought up.

  I found that the modules for the network was not being installed.
  This was due to the config CONFIG_MODULES_USE_ELF_RELA that came
  before CONFIG_MODULES, and confused ktest in thinking that
  CONFIG_MODULES=y was not found.

  Ktest needs to test all configs and not just stop if something starts
  with CONFIG_MODULES."

* tag 'ktest-v3.7-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-ktest:
  ktest: Fix ktest confusion with CONFIG_MODULES_USE_ELF_RELA

Pull minor spi MXS fixes from Mark Brown:
 "These fixes are both pretty minor ones and are driver local."

* tag 'spi-mxs' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/misc:
  spi: mxs: Terminate DMA in case of DMA timeout
  spi: mxs: Assign message status after transfer finished

Pull arm-soc fixes from Arnd Bergmann:
 "Bug fixes for a number of ARM platforms, mostly OMAP, imx and at91.

  These come a little later than I had hoped but unfortunately we had a
  few of these patches cause regressions themselves and had to work out
  how to deal with those in the meantime."

* tag 'fixes-for-3.7' of git://git.kernel.org/pub/scm/linux/kernel/git/arm/arm-soc: (38 commits)
  Revert "ARM i.MX25: Fix PWM per clock lookups"
  ARM: versatile: fix versatile_defconfig
  ARM: mvebu: update defconfig with 3.7 changes
  ARM: at91: fix at91x40 build
  ARM: socfpga: Fix socfpga compilation with early_printk() enabled
  ARM: SPEAr: Remove unused empty files
  MAINTAINERS: Add arm-soc tree entry
  ARM: dts: mxs: add the "clock-names" for gpmi-nand
  ARM: ux500: Correct SDI5 address and add some format changes
  ARM: ux500: Specify AMBA Primecell IDs for Nomadik I2C in DT
  ARM: ux500: Fix build error relating to IRQCHIP_SKIP_SET_WAKE
  ARM: at91: drop duplicated config SOC_AT91SAM9 entry
  ARM: at91/i2c: change id to let i2c-at91 work
  ARM: at91/i2c: change id to let i2c-gpio work
  ARM: at91/dts: at91sam9g20ek_common: Fix typos in buttons labels.
  ARM: at91: fix external interrupt specification in board code
  ARM: at91: fix external interrupts in non-DT case
  ARM: at91: at91sam9g10: fix SOC type detection
  ARM: at91/tc: fix typo in the DT document
  ARM: AM33XX: Fix configuration of dmtimer parent clock by dmtimer driverDate:Wed, 17 Oct 2012 13:55:55 -0500
  ...

Functions generic_file_splice_read and generic_file_splice_write access
the pagecache directly. For block devices these functions must be locked
so that block size is not changed while they are in progress.

This patch is an additional fix for commit b87570f5 ("Fix a crash
when block device is read and block size is changed at the same time")
that locked aio_read, aio_write and mmap against block size change.
Signed-off-by: NMikulas Patocka <mpatocka@redhat.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Use rcu_read_lock_sched / rcu_read_unlock_sched / synchronize_sched
instead of rcu_read_lock / rcu_read_unlock / synchronize_rcu.

This is an optimization. The RCU-protected region is very small, so
there will be no latency problems if we disable preempt in this region.

So we use rcu_read_lock_sched / rcu_read_unlock_sched that translates
to preempt_disable / preempt_disable. It is smaller (and supposedly
faster) than preemptible rcu_read_lock / rcu_read_unlock.
Signed-off-by: NMikulas Patocka <mpatocka@redhat.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This patch introduces new barrier pair light_mb() and heavy_mb() for
percpu rw semaphores.

This patch fixes a bug in percpu-rw-semaphores where a barrier was
missing in percpu_up_write.

This patch improves performance on the read path of
percpu-rw-semaphores: on non-x86 cpus, there was a smp_mb() in
percpu_up_read. This patch changes it to a compiler barrier and removes
the "#if defined(X86) ..." condition.

From: Lai Jiangshan <laijs@cn.fujitsu.com>
Signed-off-by: NMikulas Patocka <mpatocka@redhat.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This reverts commit 92063cee, it
was applied prematurely, causing this build error for
imx_v4_v5_defconfig:

arch/arm/mach-imx/clk-imx25.c: In function 'mx25_clocks_init':
arch/arm/mach-imx/clk-imx25.c:206:26: error: 'pwm_ipg_per' undeclared (first use in this function)
arch/arm/mach-imx/clk-imx25.c:206:26: note: each undeclared identifier is reported only once for each function it appears in

Sascha Hauer explains:
> There are several gates missing in clk-imx25.c. I have a patch which
> adds support for them and I seem to have missed that the above depends
> on it.
Signed-off-by: NArnd Bergmann <arnd@arndb.de>

With the introduction of CONFIG_ARCH_MULTIPLATFORM, versatile is
no longer the default platform, so we need to enable
CONFIG_ARCH_VERSATILE explicitly in order for that to be selected
rather than the multiplatform configuration.
Signed-off-by: NArnd Bergmann <arnd@arndb.de>

The split of 370 and XP into two Kconfig options and the multiplatform
kernel support has changed a few Kconfig symbols, so let's update the
mvebu_defconfig file with the latest changes.
Signed-off-by: NThomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: NArnd Bergmann <arnd@arndb.de>

patch 738a0fd7 "ARM: at91: fix external interrupts in non-DT case"
fixed a run-time error on some at91 platforms but did not apply
the same change to at91x40, which now doesn't build.

This changes at91x40 in the same way that the other platforms
were changed.
Signed-off-by: NArnd Bergmann <arnd@arndb.de>
Acked-by: NJean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com>