I noticed when filling a 1T filesystem with 4 threads using the
fs_mark benchmark:

fs_mark -d /mnt/test -D 256 -n 100000 -t 4 -s 20480 -F -S 0

that I occasionally got checksum mismatch errors:

EXT4-fs error (device sdb): ext4_init_inode_bitmap: Checksum bad for group 6935

etc.  I'd reliably get 4-5 of them during the run.

It appears that the problem is likely a race to init the bg's
when the uninit_bg feature is enabled.

With the patch below, which adds sb_bgl_locking around initialization,
I was able to complete several runs with no errors or warnings.
Signed-off-by: NEric Sandeen <sandeen@redhat.com>
Signed-off-by: NTheodore Ts'o <tytso@mit.edu>

ext4_read_block_bitmap and read_inode_bitmap do essentially
the same thing, and yet they are structured quite differently.
I came across this difference while looking at doing bg locking
during bg initialization.

This patch:

* removes unnecessary casts in the error messages
* renames read_inode_bitmap to ext4_read_inode_bitmap
* and more substantially, restructures the inode bitmap
  reading function to be more like the block bitmap counterpart.

The change to the inode bitmap reader simplifies the locking
to be applied in the next patch.
Signed-off-by: NEric Sandeen <sandeen@redhat.com>
Signed-off-by: NTheodore Ts'o <tytso@mit.edu>

This patch does block reservation for delayed
allocation, to avoid ENOSPC later at page flush time.

Blocks(data and metadata) are reserved at da_write_begin()
time, the freeblocks counter is updated by then, and the number of
reserved blocks is store in per inode counter.
        
At the writepage time, the unused reserved meta blocks are returned
back. At unlink/truncate time, reserved blocks are properly released.

Updated fix from  Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
to fix the oldallocator block reservation accounting with delalloc, added
lock to guard the counters and also fix the reservation for meta blocks.
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: NMingming Cao <cmm@us.ibm.com>
Signed-off-by: NTheodore Ts'o <tytso@mit.edu>

Delayed allocation need to check free blocks at every write time.
percpu_counter_read_positive() is not quit accurate. delayed
allocation need a more accurate accounting, but using
percpu_counter_sum_positive() is frequently is quite expensive.

This patch added a new function to update center counter when sum
per-cpu counter, to increase the accurate rate for next
percpu_counter_read() and require less calling expensive
percpu_counter_sum().
Signed-off-by: NMingming Cao <cmm@us.ibm.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

mballoc allocation missed check for blocks reserved for root users. Add
ext4_has_free_blocks() check before allocation. Also modified
ext4_has_free_blocks() to support multiple block allocation request.
Signed-off-by: NMingming Cao <cmm@us.ibm.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

Move the code related to block allocation to a single function and add helper
funtions to differient allocation for data and meta data blocks
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: NMingming Cao <cmm@us.ibm.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

When mballoc is enabled, block allocation for old block-based
files are allocated using mballoc allocator instead of old
block-based allocator. The old ext3 block reservation is turned
off when mballoc is turned on.

However, the in-core preallocation is not enabled for block-based/
non-extent based file block allocation. This result in performance
regression, as now we don't have "reservation" ore in-core preallocation
to prevent interleaved fragmentation in multiple writes workload.

This patch fix this by enable per inode in-core preallocation
for non extent files when mballoc is used.
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: NMingming Cao <cmm@us.ibm.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

When meta_bg feature is enabled and s_first_meta_bg != 0, 
ext4_init_block_bitmap() miscalculates the number of block used by
the group descriptor table (0 or 1 for metablock block group)

This patch fixes this by using ext4_bg_num_gdb()
Signed-off-by: NAkinobu Mita <akinobu.mita@gmail.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Stephen Tweedie <sct@redhat.com>
Signed-off-by: NMingming Cao <cmm@us.ibm.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>
Acked-by: NAndreas Dilger <adilger@sun.com>

Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

This patch mostly controls the way inode are allocated in order to
make ialloc aware of flex_bg block group grouping.  It achieves this
by bypassing the Orlov allocator when block group meta-data are packed
toghether through mke2fs.  Since the impact on the block allocator is
minimal, this patch should have little or no effect on other block
allocation algorithms. By controlling the inode allocation, it can
basically control where the initial search for new block begins and
thus indirectly manipulate the block allocator.

This allocator favors data and meta-data locality so the disk will
gradually be filled from block group zero upward.  This helps improve
performance by reducing seek time.  Since the group of inode tables
within one flex_bg are treated as one giant inode table, uninitialized
block groups would not need to partially initialize as many inode
table as with Orlov which would help fsck time as the filesystem usage
goes up.
Signed-off-by: NJose R. Santos <jrs@us.ibm.com>
Signed-off-by: NValerie Clement <valerie.clement@bull.net>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

Since this a non-static function, make it be ext4 specific to avoid
conflicts with potentially other filesystems.
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

if (...) BUG(); should be replaced with BUG_ON(...) when the test has no
side-effects to allow a definition of BUG_ON that drops the code completely.

The semantic patch that makes this change is as follows:
(http://www.emn.fr/x-info/coccinelle/)

// <smpl>
@ disable unlikely @ expression E,f; @@

(
  if (<... f(...) ...>) { BUG(); }
|
- if (unlikely(E)) { BUG(); }
+ BUG_ON(E);
)

@@ expression E,f; @@

(
  if (<... f(...) ...>) { BUG(); }
|
- if (E) { BUG(); }
+ BUG_ON(E);
)
// </smpl>
Signed-off-by: NJulia Lawall <julia@diku.dk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

With FLEX_BG block bitmaps, inode bitmaps and inode tables _MAY_ be
allocated outside the group.  So, when initializing an uninitialized
block bitmap, we need to check the location of this blocks before
setting the corresponding bits in the block bitmap of the newly
initialized group.  Also return the right number of free blocks when
counting the available free blocks in uninit group.
Tested-by: NAneesh Kumar K.V <aneesh.kumar@inux.vnet.ibm.com>
Signed-off-by: NJose R. Santos <jrs@us.ibm.com>
Signed-off-by: NMingming Cao <cmm@us.ibm.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

If the block allocator gets blocks out of system zone ext4 calls
ext4_error. But if the file system is mounted with errors=continue
retry block allocation. We need to mark the system zone blocks as
in use to make sure retry don't pick them again

System zone is the block range mapping block bitmap, inode bitmap and inode
table.
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: NMingming Cao <cmm@us.ibm.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

Move ext4 headers out of include/linux.  This is just the trivial move,
there's some more thing that could be done later. 
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMingming Cao <cmm@us.ibm.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

__FUNCTION__ is gcc-specific, use __func__
Signed-off-by: NHarvey Harrison <harvey.harrison@gmail.com>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

Use ext4_group_first_block_no() and assign the return values to
ext4_fsblk_t variables.
Signed-off-by: NAkinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>
Cc: Stephen Tweedie <sct@redhat.com>
Cc: adilger@clusterfs.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Mingming Cao <cmm@us.ibm.com>

replace all:
little_endian_variable = cpu_to_leX(leX_to_cpu(little_endian_variable) +
					expression_in_cpu_byteorder);
with:
	leX_add_cpu(&little_endian_variable, expression_in_cpu_byteorder);
generated with semantic patch
Signed-off-by: NMarcin Slusarz <marcin.slusarz@gmail.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>
Cc: linux-ext4@vger.kernel.org
Cc: sct@redhat.com
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: adilger@clusterfs.com
Cc: Mingming Cao <cmm@us.ibm.com>

Use ext[234]_bg_has_super() to remove duplicate code.
Signed-off-by: NAkinobu Mita <akinobu.mita@gmail.com>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The comment in ext[234]_new_blocks() describes about "i".  But there is no
local variable called "i" in that scope.  I guess it has been renamed to
group_no.

Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: NAkinobu Mita <akinobu.mita@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: NAlex Tomas <alex@clusterfs.com>
Signed-off-by: NAndreas Dilger <adilger@clusterfs.com>
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: NEric Sandeen <sandeen@redhat.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

Builds with EXT4FS_DEBUG defined (to enable ext4_debug()) fail
without these changes.  Clean up some format warnings too.
Signed-off-by: NEric Sandeen <sandeen@redhat.com>
Signed-off-by: NMingming Cao <cmm@us.ibm.com>

We are currently taking the truncate_mutex for every read. This would have
performance impact on large CPU configuration. Convert the lock to read write
semaphore and take read lock when we are trying to read the file.
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>

When a new block bitmap is read from disk in read_block_bitmap()
there are a few bits that should ALWAYS be set.  In particular,
the blocks given corresponding to block bitmap, inode bitmap and inode tables.
Validate the block bitmap against these blocks.
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>

This fix some instances where we were continuing after calling
ext4_error. ext4_error call panic only if errors=panic mount option is
set. So we need to make sure we return correctly after ext4_error call

Reported by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>

In many places variables for block group are of type int, which limits the
maximum number of block groups to 2^31.  Each block group can have up to
2^15 blocks, with a 4K block size,  and the max filesystem size is limited to
2^31 * (2^15 * 2^12) = 2^58  -- or 256 PB

This patch introduces a new type ext4_group_t, of type unsigned long, to
represent block group numbers in ext4.
All occurrences of block group variables are converted to type ext4_group_t.
Signed-off-by: NAvantika Mathur <mathur@us.ibm.com>

This reverts commit 7c9e69fa, fixing up
conflicts in fs/ext4/balloc.c manually.

The cost of doing the bitmap validation on each lookup - even when the
bitmap is cached - is absolutely prohibitive.  We could, and probably
should, do it only when adding the bitmap to the buffer cache.  However,
right now we are better off just reverting it.

Peter Zijlstra measured the cost of this extra validation as a 85%
decrease in cached iozone, and while I had a patch that took it down to
just 17% by not being _quite_ so stupid in the validation, it was still
a big slowdown that could have been avoided by just doing it right.

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Aneesh Kumar <aneesh.kumar@linux.vnet.ibm.com>
Cc: Andreas Dilger <adilger@clusterfs.com>
Cc: Mingming Cao <cmm@us.ibm.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Convert bg_inode_bitmap and bg_inode_table to bg_inode_bitmap_lo
and bg_inode_table_lo.  This helps in finding BUGs due to
direct partial access of these split 64 bit values

Also fix one direct partial access
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>

In pass1 of e2fsck, every inode table in the fileystem is scanned and checked,
regardless of whether it is in use.  This is this the most time consuming part
of the filesystem check.  The unintialized block group feature can greatly
reduce e2fsck time by eliminating checking of uninitialized inodes.

With this feature, there is a a high water mark of used inodes for each block
group.  Block and inode bitmaps can be uninitialized on disk via a flag in the
group descriptor to avoid reading or scanning them at e2fsck time.  A checksum
of each group descriptor is used to ensure that corruption in the group
descriptor's bit flags does not cause incorrect operation.

The feature is enabled through a mkfs option

	mke2fs /dev/ -O uninit_groups

A patch adding support for uninitialized block groups to e2fsprogs tools has
been posted to the linux-ext4 mailing list.

The patches have been stress tested with fsstress and fsx.  In performance
tests testing e2fsck time, we have seen that e2fsck time on ext3 grows
linearly with the total number of inodes in the filesytem.  In ext4 with the
uninitialized block groups feature, the e2fsck time is constant, based
solely on the number of used inodes rather than the total inode count.
Since typical ext4 filesystems only use 1-10% of their inodes, this feature can
greatly reduce e2fsck time for users.  With performance improvement of 2-20
times, depending on how full the filesystem is.

The attached graph shows the major improvements in e2fsck times in filesystems
with a large total inode count, but few inodes in use.

In each group descriptor if we have

EXT4_BG_INODE_UNINIT set in bg_flags:
        Inode table is not initialized/used in this group. So we can skip
        the consistency check during fsck.
EXT4_BG_BLOCK_UNINIT set in bg_flags:
        No block in the group is used. So we can skip the block bitmap
        verification for this group.

We also add two new fields to group descriptor as a part of
uninitialized group patch.

        __le16  bg_itable_unused;       /* Unused inodes count */
        __le16  bg_checksum;            /* crc16(sb_uuid+group+desc) */

bg_itable_unused:

If we have EXT4_BG_INODE_UNINIT not set in bg_flags
then bg_itable_unused will give the offset within
the inode table till the inodes are used. This can be
used by fsck to skip list of inodes that are marked unused.

bg_checksum:
Now that we depend on bg_flags and bg_itable_unused to determine
the block and inode usage, we need to make sure group descriptor
is not corrupt. We add checksum to group descriptor to
detect corruption. If the descriptor is found to be corrupt, we
mark all the blocks and inodes in the group used.
Signed-off-by: NAvantika Mathur <mathur@us.ibm.com>
Signed-off-by: NAndreas Dilger <adilger@clusterfs.com>
Signed-off-by: NMingming Cao <cmm@us.ibm.com>
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>

When a new block bitmap is read from disk in read_block_bitmap() there are
a few bits that should ALWAYS be set.  In particular, the blocks given by
ext4_blk_bitmap, ext4_inode_bitmap and ext4_inode_table.  Validate the
block bitmap against these blocks.

[akpm@linux-foundation.org: cleanups]
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: NAndreas Dilger <adilger@clusterfs.com>
Acked-by: NMingming Cao <cmm@us.ibm.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Hugh spotted that some code does:
  percpu_counter_add(&counter, -unsignedlong)

which, when the amount argument is of type s32, sort-of works thanks to
two's-complement. However when we'd change the type to s64 this breaks on 32bit
machines, because the promotion rules zero extend the unsigned number.

Provide percpu_counter_sub() to hide the s64 cast. That is:
  percpu_counter_sub(&counter, foo)
is equal to:
  percpu_counter_add(&counter, -(s64)foo);
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Hugh Dickins <hugh@veritas.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

 s/percpu_counter_mod/percpu_counter_add/

Because its a better name, _mod implies modulo.
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

When the JBD code was forked to create the new JBD2 code base, the
references to CONFIG_JBD_DEBUG where never changed to
CONFIG_JBD2_DEBUG.  This patch fixes that.
Signed-off-by: NJose R. Santos <jrs@us.ibm.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

In ext4_new_blocks(), one of two ext4_block_bitmap() calls should be
ext4_inode_bitmap() call.  It is not harmful in normal processing, but it
should be fixed.
Signed-off-by: NToshiyuki Okajima <toshi.okajima@jp.fujitsu.com>
Cc: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Replace a lot of spaces with tabs
Signed-off-by: NDave Kleikamp <shaggy@linux.vnet.ibm.com>
Signed-off-by: N"Theodore Ts'o" <tytso@mit.edu>

Signed-off-by: N"Aneesh Kumar K.V" <aneesh.kumar@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Port fix to the off-by-one in find_next_usable_block's memscan from ext2 to
ext4; but it didn't cause a serious problem for ext4 because the additional
ext4_test_allocatable check rescued it from the error.

[akpm@osdl.org: build fix]
Signed-off-by: NMingming Cao <cmm@us.ibm.com>
Signed-off-by: NHugh Dickins <hugh@veritas.com>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

ext4_new_blocks has a nice io_error label for setting -EIO, so goto that in
the one place that doesn't already use it.
Signed-off-by: NMingming Cao <cmm@us.ibm.com>
Signed-off-by: NHugh Dickins <hugh@veritas.com>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

The reservations tree is an rb_tree not a list, so it's less confusing to use
rb_entry() than list_entry() - though they're both just container_of().
Signed-off-by: NMingming Cao <cmm@us.ibm.com>
Signed-off-by: NHugh Dickins <hugh@veritas.com>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

rsv_end is the last block within the reservation, so alloc_new_reservation
should accept start_block == rsv_end as success.
Signed-off-by: NMingming Cao <cmm@us.ibm.com>
Signed-off-by: NHugh Dickins <hugh@veritas.com>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>