openwrt/staging/blogic.git
15 years agojbd2: Fix oops in jbd2_journal_init_inode() on corrupted fs
Jan Kara [Tue, 6 Jan 2009 19:53:35 +0000 (14:53 -0500)]
jbd2: Fix oops in jbd2_journal_init_inode() on corrupted fs

On 32-bit system with CONFIG_LBD getblk can fail because provided
block number is too big.  Add error checks so we fail gracefully if
getblk() returns NULL (which can also happen on memory allocation
failures).

Thanks to David Maciejak from Fortinet's FortiGuard Global Security
Research Team for reporting this bug.

http://bugzilla.kernel.org/show_bug.cgi?id=12370

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
cc: stable@kernel.org

15 years agoext4: Remove "extents" mount option
Theodore Ts'o [Tue, 6 Jan 2009 19:53:16 +0000 (14:53 -0500)]
ext4: Remove "extents" mount option

This mount option is largely superfluous, and in fact the way it was
implemented was buggy; if a filesystem which did not have the extents
feature flag was mounted -o extents, the filesystem would attempt to
create and use extents-based file even though the extents feature flag
was not eabled.  The simplest thing to do is to nuke the mount option
entirely.  It's not all that useful to force the non-creation of new
extent-based files if the filesystem can support it.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoblock: Add Kconfig help which notes that ext4 needs CONFIG_LBD
Theodore Ts'o [Tue, 6 Jan 2009 20:16:33 +0000 (15:16 -0500)]
block: Add Kconfig help which notes that ext4 needs CONFIG_LBD

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Jens Axboe <jens.axboe@oracle.com>
15 years agoext4: Make printk's consistently prefixed with "EXT4-fs: "
Theodore Ts'o [Tue, 6 Jan 2009 05:20:32 +0000 (00:20 -0500)]
ext4: Make printk's consistently prefixed with "EXT4-fs: "

Previously, some were "ext4: ", and some were "EXT4: "; change them to
be consistent with most ext4 printk's, which is to use "EXT4-fs: ".

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: Add sanity checks for the superblock before mounting the filesystem
Theodore Ts'o [Tue, 6 Jan 2009 19:53:26 +0000 (14:53 -0500)]
ext4: Add sanity checks for the superblock before mounting the filesystem

This avoids insane superblock configurations that could lead to kernel
oops due to null pointer derefences.

http://bugzilla.kernel.org/show_bug.cgi?id=12371

Thanks to David Maciejak at Fortinet's FortiGuard Global Security
Research Team who discovered this bug independently (but at
approximately the same time) as Thiemo Nagel, who submitted the patch.

Signed-off-by: Thiemo Nagel <thiemo.nagel@ph.tum.de>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@kernel.org
15 years agoext4: Add mount option to set kjournald's I/O priority
Theodore Ts'o [Tue, 6 Jan 2009 03:46:26 +0000 (22:46 -0500)]
ext4: Add mount option to set kjournald's I/O priority

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Jens Axboe <jens.axboe@oracle.com>
15 years agojbd2: Submit writes to the journal using WRITE_SYNC
Theodore Ts'o [Mon, 5 Jan 2009 00:55:57 +0000 (19:55 -0500)]
jbd2: Submit writes to the journal using WRITE_SYNC

Since we will be waiting the write of the commit record to the journal
to complete in journal_submit_commit_record(), submit it using
WRITE_SYNC.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agojbd2: Add pid and journal device name to the "kjournald2 starting" message
Theodore Ts'o [Sun, 4 Jan 2009 03:56:44 +0000 (22:56 -0500)]
jbd2: Add pid and journal device name to the "kjournald2 starting" message

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: Add markers for better debuggability
Theodore Ts'o [Sun, 4 Jan 2009 01:03:21 +0000 (20:03 -0500)]
ext4: Add markers for better debuggability

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: Remove code to create the journal inode
Theodore Ts'o [Tue, 6 Jan 2009 16:14:25 +0000 (11:14 -0500)]
ext4: Remove code to create the journal inode

This code has been obsolete in quite some time, since the supported
method for adding a journal inode is to use tune2fs (or to creating
new filesystem with a journal via mke2fs or mkfs.ext4).

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: provide function to release metadata pages under memory pressure
Toshiyuki Okajima [Tue, 6 Jan 2009 03:38:48 +0000 (22:38 -0500)]
ext4: provide function to release metadata pages under memory pressure

Pages in the page cache belonging to ext4 data files are released via
the ext4_releasepage() function specified in the ext4 inode's
address_space_ops.  However, metadata blocks (such as indirect blocks,
directory blocks, etc) are managed via the block device
address_space_ops, and they can not be released by
try_to_free_buffers() if they have a journal head attached to them.

To address this, we supply a release_metadata function which calls
jbd2_journal_try_to_free_buffers() function to free the metadata, and
which is called by the block device's blkdev_releasepage() function.

Signed-off-by: Toshiyuki Okajima <toshi.okajima@jp.fujitsu.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: linux-fsdevel@vger.kernel.org
15 years agoext3: provide function to release metadata pages under memory pressure
Toshiyuki Okajima [Tue, 6 Jan 2009 03:38:14 +0000 (22:38 -0500)]
ext3: provide function to release metadata pages under memory pressure

Pages in the page cache belonging to ext3 data files are released via
the ext3_releasepage() function specified in the ext3 inode's
address_space_ops.  However, metadata blocks (such as indirect blocks,
directory blocks, etc) are managed via the block device
address_space_ops, and they can not be released by
try_to_free_buffers() if they have a journal head attached to them.

To address this, we supply a try_to_free_pages() function which calls
journal_try_to_free_buffers() function to free the metadata, and which
is called by the block device's blkdev_releasepage() function.

Signed-off-by: Toshiyuki Okajima <toshi.okajima@jp.fujitsu.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: linux-fsdevel@vger.kernel.org
15 years agoadd releasepage hooks to block devices which can be used by file systems
Theodore Ts'o [Sat, 3 Jan 2009 14:47:09 +0000 (09:47 -0500)]
add releasepage hooks to block devices which can be used by file systems

Implement blkdev_releasepage() to release the buffer_heads and pages
after we release private data belonging to a mounted filesystem.

Cc: Toshiyuki Okajima <toshi.okajima@jp.fujitsu.com>
Cc: linux-fsdevel@vger.kernel.org
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: Fix s_dirty_blocks_counter if block allocation failed with nodelalloc
Aneesh Kumar K.V [Tue, 6 Jan 2009 02:49:12 +0000 (21:49 -0500)]
ext4: Fix s_dirty_blocks_counter if block allocation failed with nodelalloc

With nodelalloc option we need to update the dirty block counter on
block allocation failure. This is needed because we increment the
dirty block counter early in the block allocation phase. Without
the patch s_dirty_blocks_counter goes wrong so that filesystem's
free blocks decreases incorrectly.

Tested-by: Akira Fujita <a-fujita@rs.jp.nec.com>
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@kernel.org
15 years agoext4: Init the complete page while building buddy cache
Aneesh Kumar K.V [Tue, 6 Jan 2009 02:48:56 +0000 (21:48 -0500)]
ext4: Init the complete page while building buddy cache

We need to init the complete page during buddy cache init
by setting the contents to '1'.  Otherwise we can see the
following errors after doing an online resize of the
filesystem:

EXT4-fs error (device sdb1): ext4_mb_mark_diskspace_used:
Allocating block 1040385 in system zone of 127 group

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@kernel.org
15 years agoext4: Don't allow new groups to be added during block allocation
Aneesh Kumar K.V [Tue, 6 Jan 2009 02:46:55 +0000 (21:46 -0500)]
ext4: Don't allow new groups to be added during block allocation

After we mark the blocks in the buddy cache as allocated,
we need to ensure that we don't reinit the buddy cache until
the block bitmap is updated.  This commit achieves this by holding
the group_info alloc_semaphore till ext4_mb_release_context

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@kernel.org
15 years agoext4: mark the blocks/inode bitmap beyond end of group as used
Aneesh Kumar K.V [Tue, 6 Jan 2009 02:46:04 +0000 (21:46 -0500)]
ext4: mark the blocks/inode bitmap beyond end of group as used

We need to mark the block/inode bitmap beyond the end of the group
with '1'.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@kernel.org
15 years agoext4: Use new buffer_head flag to check uninit group bitmaps initialization
Aneesh Kumar K.V [Tue, 6 Jan 2009 02:49:55 +0000 (21:49 -0500)]
ext4: Use new buffer_head flag to check uninit group bitmaps initialization

For uninit block group, the on-disk bitmap is not initialized. That
implies we cannot depend on the uptodate flag on the bitmap
buffer_head to find bitmap validity.  Use a new buffer_head flag which
would be set after we properly initialize the bitmap.  This also
prevents (re-)initializing the uninit group bitmap every time we call
ext4_read_block_bitmap().

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@kernel.org
15 years agoext4: Fix the race between read_inode_bitmap() and ext4_new_inode()
Aneesh Kumar K.V [Tue, 6 Jan 2009 02:38:14 +0000 (21:38 -0500)]
ext4: Fix the race between read_inode_bitmap() and ext4_new_inode()

We need to make sure we update the inode bitmap and clear
EXT4_BG_INODE_UNINIT flag with sb_bgl_lock held, since
ext4_read_inode_bitmap() looks at EXT4_BG_INODE_UNINIT to decide
whether to initialize the inode bitmap each time it is called.
(introduced by commit c806e68f.)

ext4_read_inode_bitmap does:

spin_lock(sb_bgl_lock(EXT4_SB(sb), block_group));
if (desc->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT)) {
ext4_init_inode_bitmap(sb, bh, block_group, desc);

and ext4_new_inode does
if (!ext4_set_bit_atomic(sb_bgl_lock(sbi, group),
                   ino, inode_bitmap_bh->b_data))
   ......
   ...
spin_lock(sb_bgl_lock(sbi, group));

gdp->bg_flags &= cpu_to_le16(~EXT4_BG_INODE_UNINIT);
i.e., on allocation we update the bitmap then we take the sb_bgl_lock
and clear the EXT4_BG_INODE_UNINIT flag. What can happen is a
parallel ext4_read_inode_bitmap can zero out the bitmap in between
the above ext4_set_bit_atomic and spin_lock(sb_bg_lock..)

The race results in below user visible errors
EXT4-fs error (device sdb1): ext4_free_inode: bit already cleared for inode 168449
EXT4-fs warning (device sdb1): ext4_unlink: Deleting nonexistent file ...
EXT4-fs warning (device sdb1): ext4_rmdir: empty directory has too many links ...
# ls -al /mnt/tmp/f/p369/d3/d6/d39/db2/dee/d10f/d3f/l71
ls: /mnt/tmp/f/p369/d3/d6/d39/db2/dee/d10f/d3f/l71: Stale NFS file handle

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@kernel.org
15 years agoext4: code cleanup
Aneesh Kumar K.V [Sun, 4 Jan 2009 03:33:39 +0000 (22:33 -0500)]
ext4: code cleanup

Rename some variables.  We also unlock locks in the reverse order we
acquired as a part of cleanup.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: Use high 16 bits of the block group descriptor's free counts fields
Aneesh Kumar K.V [Tue, 6 Jan 2009 03:20:24 +0000 (22:20 -0500)]
ext4: Use high 16 bits of the block group descriptor's free counts fields

Rename the lower bits with suffix _lo and add helper
to access the values. Also rename bg_itable_unused_hi
to bg_pad as in e2fsprogs.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: Fix race between read_block_bitmap() and mark_diskspace_used()
Aneesh Kumar K.V [Tue, 6 Jan 2009 02:38:26 +0000 (21:38 -0500)]
ext4: Fix race between read_block_bitmap() and mark_diskspace_used()

We need to make sure we update the block bitmap and clear
EXT4_BG_BLOCK_UNINIT flag with sb_bgl_lock held, since
ext4_read_block_bitmap() looks at EXT4_BG_BLOCK_UNINIT to decide
whether to initialize the block bitmap each time it is called
(introduced by commit c806e68f), and this can race with block
allocations in ext4_mb_mark_diskspace_used().

ext4_read_block_bitmap does:

spin_lock(sb_bgl_lock(EXT4_SB(sb), block_group));
if (desc->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
ext4_init_block_bitmap(sb, bh, block_group, desc);

Now on the block allocation side we do

mb_set_bits(sb_bgl_lock(sbi, ac->ac_b_ex.fe_group), bitmap_bh->b_data,
ac->ac_b_ex.fe_start, ac->ac_b_ex.fe_len);
....
spin_lock(sb_bgl_lock(sbi, ac->ac_b_ex.fe_group));
if (gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) {
gdp->bg_flags &= cpu_to_le16(~EXT4_BG_BLOCK_UNINIT);

ie on allocation we update the bitmap then we take the sb_bgl_lock
and clear the EXT4_BG_BLOCK_UNINIT flag. What can happen is a
parallel ext4_read_block_bitmap can zero out the bitmap in between
the above mb_set_bits and spin_lock(sb_bg_lock..)

The race results in below user visible errors
EXT4-fs error (device sdb1): ext4_mb_release_inode_pa: free 100, pa_free 105
EXT4-fs error (device sdb1): mb_free_blocks: double-free of inode 0's block ..

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@kernel.org
15 years agoext4: fix BUG when calling ext4_error with locked block group
Aneesh Kumar K.V [Tue, 6 Jan 2009 03:19:52 +0000 (22:19 -0500)]
ext4: fix BUG when calling ext4_error with locked block group

The mballoc code likes to call ext4_error while it is holding locked
block groups.  This can causes a scheduling in atomic context BUG.  We
can't just unlock the block group and relock it after/if ext4_error
returns since that might result in race conditions in the case where
the filesystem is set to continue after finding errors.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoext4: Fix lockdep recursive locking warning
Aneesh Kumar K.V [Mon, 24 Nov 2008 04:51:53 +0000 (23:51 -0500)]
ext4: Fix lockdep recursive locking warning

In ext4_mb_init_group(), if the filesystem block size is less than
PAGE_SIZE/2, the code tries to grab alloc_sem for multiple block
groups in a loop.  We need to allow for this by using
down_write_nested() and passing in the loop index as a lock subclass
number.  This works because no other code path needs to take multiple
alloc_sem's.  Note that lockdep will fail for filesystem blocksize
smaller than to PAGE_SIZE/16k.  (e.g., a 1k filesystem blocksize with
a 32k page size, or a 2k filesystem blocksize with a 64k blocksize,
etc.)

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: don't use blocks freed but not yet committed in buddy cache init
Aneesh Kumar K.V [Tue, 6 Jan 2009 02:36:55 +0000 (21:36 -0500)]
ext4: don't use blocks freed but not yet committed in buddy cache init

When we generate buddy cache (especially during resize) we need to
make sure we don't use the blocks freed but not yet comitted.  This
makes sure we have the right value of free blocks count in the group
info and also in the bitmap.  This also ensures the ordered mode
consistency

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@kernel.org
16 years agojbd2: Call journal commit callback without holding j_list_lock
Aneesh Kumar K.V [Thu, 6 Nov 2008 22:50:21 +0000 (17:50 -0500)]
jbd2: Call journal commit callback without holding j_list_lock

Avoid freeing the transaction in __jbd2_journal_drop_transaction() so
the journal commit callback can run without holding j_list_lock, to
avoid lock contention on this spinlock.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoext4: cleanup mballoc header files
Aneesh Kumar K.V [Tue, 25 Nov 2008 20:11:52 +0000 (15:11 -0500)]
ext4: cleanup mballoc header files

Move some of the forward declaration of the static functions
to mballoc.c where they are used. This enables us to include
mballoc.h in other .c files. Also correct the buddy cache
documentation.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: Use EXT4_GROUP_INFO_NEED_INIT_BIT during resize
Aneesh Kumar K.V [Tue, 6 Jan 2009 02:36:19 +0000 (21:36 -0500)]
ext4: Use EXT4_GROUP_INFO_NEED_INIT_BIT during resize

The new groups added during resize are flagged as
need_init group. Make sure we properly initialize these
groups. When we have block size < page size and we are adding
new groups the page may still be marked uptodate even though
we haven't initialized the group. While forcing the init
of buddy cache we need to make sure other groups part of the
same page of buddy cache is not using the cache.
group_info->alloc_sem is added to ensure the same.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
cc: stable@kernel.org

15 years agoext4: Add blocks added during resize to bitmap
Aneesh Kumar K.V [Tue, 6 Jan 2009 02:36:02 +0000 (21:36 -0500)]
ext4: Add blocks added during resize to bitmap

With this change new blocks added during resize
are marked as free in the block bitmap and the
group is flagged with EXT4_GROUP_INFO_NEED_INIT_BIT
flag.  This makes sure when mballoc tries to allocate
blocks from the new group we would reload the
buddy information using the bitmap present in the disk.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@kernel.org
16 years agoext4: sparse fixes
Aneesh Kumar K.V [Sat, 22 Nov 2008 20:04:59 +0000 (15:04 -0500)]
ext4: sparse fixes

* Change EXT4_HAS_*_FEATURE to return a boolean
* Add a function prototype for ext4_fiemap() in ext4.h
* Make ext4_ext_fiemap_cb() and ext4_xattr_fiemap() be static functions
* Add lock annotations to mb_free_blocks()

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agojbd2: Remove a large array of bh's from the stack of the checkpoint routine
Theodore Ts'o [Wed, 5 Nov 2008 05:09:22 +0000 (00:09 -0500)]
jbd2: Remove a large array of bh's from the stack of the checkpoint routine

jbd2_log_do_checkpoint()n is one of the kernel's largest stack users.
Move the array of buffer head's from the stack of jbd2_log_do_checkpoint()
to the in-core journal structure.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoext4: Change unsigned long to unsigned int
Theodore Ts'o [Wed, 5 Nov 2008 05:14:04 +0000 (00:14 -0500)]
ext4: Change unsigned long to unsigned int

Convert the unsigned longs that are most responsible for bloating the
stack usage on 64-bit systems.

Nearly all places in the ext3/4 code which uses "unsigned long" is
probably a bug, since on 32-bit systems a ulong a 32-bits, which means
we are wasting stack space on 64-bit systems.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: Make ext4_group_t be an unsigned int
Theodore Ts'o [Tue, 6 Jan 2009 03:18:16 +0000 (22:18 -0500)]
ext4: Make ext4_group_t be an unsigned int

Nearly all places in the ext3/4 code which uses "unsigned long" is
probably a bug, since on 32-bit systems a ulong a 32-bits, which means
we are wasting stack space on 64-bit systems.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoext4: Remove i_ext_generation from ext4_inode_info structure
Theodore Ts'o [Tue, 4 Nov 2008 23:46:03 +0000 (18:46 -0500)]
ext4: Remove i_ext_generation from ext4_inode_info structure

The i_ext_generation was incremented, but never used.  Remove it to
slim down the ext4_inode_info structure.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: add fsync batch tuning knobs
Theodore Ts'o [Sun, 4 Jan 2009 01:27:38 +0000 (20:27 -0500)]
ext4: add fsync batch tuning knobs

Add new mount options, min_batch_time and max_batch_time, which
controls how long the jbd2 layer should wait for additional filesystem
operations to get batched with a synchronous write transaction.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: display average commit time
Theodore Ts'o [Wed, 17 Dec 2008 05:20:45 +0000 (00:20 -0500)]
ext4: display average commit time

Display the average commit time (which is used by the ext4 fsync
batching patch) in /proc/fs/jbd2/*/info for performance tuning
purposes.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agojbd2: improve jbd2 fsync batching
Josef Bacik [Wed, 26 Nov 2008 06:14:26 +0000 (01:14 -0500)]
jbd2: improve jbd2 fsync batching

This patch removes the static sleep time in favor of a more self
optimizing approach where we measure the average amount of time it
takes to commit a transaction to disk and the ammount of time a
transaction has been running.  If somebody does a sync write or an
fsync() traditionally we would sleep for 1 jiffies, which depending on
the value of HZ could be a significant amount of time compared to how
long it takes to commit a transaction to the underlying storage.  With
this patch instead of sleeping for a jiffie, we check to see if the
amount of time this transaction has been running is less than the
average commit time, and if it is we sleep for the delta using
schedule_hrtimeout to give us a higher precision sleep time.  This
greatly benefits high end storage where you could end up sleeping for
longer than it takes to commit the transaction and therefore sitting
idle instead of allowing the transaction to be committed by keeping
the sleep time to a minimum so you are sure to always be doing
something.

Signed-off-by: Josef Bacik <jbacik@redhat.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: Don't overwrite allocation_context ac_status
Aneesh Kumar K.V [Tue, 6 Jan 2009 02:34:30 +0000 (21:34 -0500)]
ext4: Don't overwrite allocation_context ac_status

We can call ext4_mb_check_limits even after successfully allocating
the requested blocks.  In that case, make sure we don't overwrite
ac_status if it already has the status AC_STATUS_FOUND.  This fixes
the lockdep warning:

=============================================
[ INFO: possible recursive locking detected ]
2.6.28-rc6-autokern1 #1
---------------------------------------------
fsstress/11948 is trying to acquire lock:
 (&meta_group_info[i]->alloc_sem){----}, at: [<c04d9a49>] ext4_mb_load_buddy+0x9f/0x278
.....

stack backtrace:
.....
 [<c04db974>] ext4_mb_regular_allocator+0xbb5/0xd44
.....

but task is already holding lock:
 (&meta_group_info[i]->alloc_sem){----}, at: [<c04d9a49>] ext4_mb_load_buddy+0x9f/0x278

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@kernel.org
15 years agoext4: remove extraneous newlines from calls to ext4_error() and ext4_warning()
Theodore Ts'o [Tue, 6 Jan 2009 03:17:35 +0000 (22:17 -0500)]
ext4: remove extraneous newlines from calls to ext4_error() and ext4_warning()

This removes annoying blank syslog entries emitted by ext4_error() or
ext4_warning(), since these functions add their own newline.

Signed-off-by: Nick Warne <nick@ukfsn.org>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agojbd2: Add barrier not supported test to journal_wait_on_commit_record
Theodore Ts'o [Tue, 6 Jan 2009 02:34:13 +0000 (21:34 -0500)]
jbd2: Add barrier not supported test to journal_wait_on_commit_record

Xen doesn't report that barriers are not supported until buffer I/O is
reported as completed, instead of when the buffer I/O is submitted.
Add a check and a fallback codepath to journal_wait_on_commit_record()
to detect this case, so that attempts to mount ext4 filesystems on
LVM/devicemapper devices on Xen guests don't blow up with an "Aborting
journal on device XXX"; "Remounting filesystem read-only" error.

Thanks to Andreas Sundstrom for reporting this issue.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@kernel.org
15 years agoext4: Allow ext4 to run without a journal
Frank Mayhar [Wed, 7 Jan 2009 05:06:22 +0000 (00:06 -0500)]
ext4: Allow ext4 to run without a journal

A few weeks ago I posted a patch for discussion that allowed ext4 to run
without a journal.  Since that time I've integrated the excellent
comments from Andreas and fixed several serious bugs.  We're currently
running with this patch and generating some performance numbers against
both ext2 (with backported reservations code) and ext4 with and without
a journal.  It just so happens that running without a journal is
slightly faster for most everything.

We did
iozone -T -t 4 s 2g -r 256k -T -I -i0 -i1 -i2

which creates 4 threads, each of which create and do reads and writes on
a 2G file, with a buffer size of 256K, using O_DIRECT for all file opens
to bypass the page cache.  Results:

                     ext2        ext4, default   ext4, no journal
  initial writes   13.0 MB/s        15.4 MB/s          15.7 MB/s
  rewrites         13.1 MB/s        15.6 MB/s          15.9 MB/s
  reads            15.2 MB/s        16.9 MB/s          17.2 MB/s
  re-reads         15.3 MB/s        16.9 MB/s          17.2 MB/s
  random readers    5.6 MB/s         5.6 MB/s           5.7 MB/s
  random writers    5.1 MB/s         5.3 MB/s           5.4 MB/s

So it seems that, so far, this was a useful exercise.

Signed-off-by: Frank Mayhar <fmayhar@google.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: Widen type of ext4_sb_info.s_mb_maxs[]
Yasunori Goto [Wed, 17 Dec 2008 05:48:39 +0000 (00:48 -0500)]
ext4: Widen type of ext4_sb_info.s_mb_maxs[]

I chased the cause of following ext4 oops report which is tested on
ia64 box.

http://bugzilla.kernel.org/show_bug.cgi?id=12018

The cause is the size of s_mb_maxs array that is defined as "unsigned
short" in ext4_sb_info structure.  If the file system's block size is
8k or greater, an unsigned short is not wide enough to contain the
value fs->blocksize << 3.

Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Miao Xie <miaox@cn.fujitsu.com>
Cc: stable@kernel.org
16 years agoext4: When resizing set the EXT4_BG_INODE_ZEROED flag for new block groups
Solofo.Ramangalahy@bull.net [Thu, 27 Nov 2008 04:44:10 +0000 (23:44 -0500)]
ext4: When resizing set the EXT4_BG_INODE_ZEROED flag for new block groups

The inode table has been zeroed in setup_new_group_blocks().  Mark it as
such in ext4_group_add().  Since we are currently clearing inode table
for the new block group, we should set the EXT4_BG_INODE_ZEROED flag.
If at some point in the future we don't immediately zero out the inode
table as part of the resize operation, then obviously we shouldn't do
this.

Signed-off-by: Solofo.Ramangalahy@bull.net
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoext4: Use simple_strtol() instead of simple_strtoul() in ext4_ui_proc_open
Roel Kluin [Wed, 26 Nov 2008 07:23:19 +0000 (02:23 -0500)]
ext4: Use simple_strtol() instead of simple_strtoul() in ext4_ui_proc_open

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agojbd2: Add BH_JBDPrivateStart
Mark Fasheh [Tue, 25 Nov 2008 22:42:31 +0000 (17:42 -0500)]
jbd2: Add BH_JBDPrivateStart

Add this so that file systems using JBD2 can safely allocate unused b_state
bits.

In this case, we add it so that Ocfs2 can define a single bit for tracking
the validation state of a buffer.

Signed-off-by: Mark Fasheh <mfasheh@suse.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoext4: fix build warning
Wu Fengguang [Tue, 25 Nov 2008 22:24:23 +0000 (17:24 -0500)]
ext4: fix build warning

Replace `if' with `goto' to assure gcc that ix has been initialized.

Signed-off-by: Wu Fengguang <wfg@linux.intel.com>
15 years agoext4: avoid ext4_error when mounting a fs with a single bg
Aneesh Kumar K.V [Tue, 6 Jan 2009 02:51:07 +0000 (21:51 -0500)]
ext4: avoid ext4_error when mounting a fs with a single bg

Remove some completely unneeded code which which caused an ext4_error
to be generated when mounting a file system with only a single block
group.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@kernel.org
15 years agoext4: Fix the delalloc writepages to allocate blocks at the right offset.
Aneesh Kumar K.V [Tue, 6 Jan 2009 02:50:43 +0000 (21:50 -0500)]
ext4: Fix the delalloc writepages to allocate blocks at the right offset.

When iterating through the pages which have mapped buffer_heads, we
failed to update the b_state value. This results in allocating blocks
at logical offset 0.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@kernel.org
16 years agoext4: tone down ext4_da_writepages warnings
Theodore Ts'o [Wed, 5 Nov 2008 14:22:24 +0000 (09:22 -0500)]
ext4: tone down ext4_da_writepages warnings

If the filesystem has errors, ext4_da_writepages() will return a *lot*
of errors, including lots and lots of stack dumps.  While it's true
that we are dropping user data on the floor, which is unfortunate, the
stack dumps aren't helpful, and they tend to obscure the true original
root cause of the problem.  So in the case where the filesystem has
aborted, return an EROFS right away.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: remove do_blk_alloc()
Theodore Ts'o [Fri, 12 Dec 2008 17:41:28 +0000 (12:41 -0500)]
ext4: remove do_blk_alloc()

The convenience function do_blk_alloc() is a static function with only
one caller, so fold it into ext4_new_meta_blocks() to simplify the
code and to make it easier to understand.

To save more stack space, if count is a null pointer in
ext4_new_meta_blocks() assume that caller wanted a single block (and
if there is an error, no blocks were allocated).

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: remove ext4_new_meta_block()
Theodore Ts'o [Sun, 7 Dec 2008 19:10:54 +0000 (14:10 -0500)]
ext4: remove ext4_new_meta_block()

There were only two one callers of the function ext4_new_meta_block(),
which just a very simpler wrapper function around
ext4_new_meta_blocks().  Change those two functions to call
ext4_new_meta_blocks() directly, to save code and stack space usage.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext4: remove ext4_new_blocks() and call ext4_mb_new_blocks() directly
Theodore Ts'o [Fri, 2 Jan 2009 04:59:43 +0000 (23:59 -0500)]
ext4: remove ext4_new_blocks() and call ext4_mb_new_blocks() directly

There was only one caller of the compatibility function
ext4_new_blocks(), in balloc.c's ext4_alloc_blocks().  Change it to
call ext4_mb_new_blocks() directly, and remove ext4_new_blocks()
altogether.  This cleans up the code, by removing two extra functions
from the call chain, and hopefully saving some stack usage.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoUpdate Documentation/filesystems/ext4.txt
Theodore Ts'o [Tue, 6 Jan 2009 19:53:06 +0000 (14:53 -0500)]
Update Documentation/filesystems/ext4.txt

Fix paragraph with recommendations on how to tune ext4 for benchmarks.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
15 years agoext3/4: Fix loop index in do_split() so it is signed
Theodore Ts'o [Sat, 6 Dec 2008 21:58:39 +0000 (16:58 -0500)]
ext3/4: Fix loop index in do_split() so it is signed

This fixes a gcc warning but it doesn't appear able to result in a
failure, since the primary way the loop is exited is the first
conditional in the for loop, and at least for a consistent filesystem,
the signed/unsigned should in practice never be exposed.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoext4: Add support for non-native signed/unsigned htree hash algorithms
Theodore Ts'o [Tue, 28 Oct 2008 17:21:44 +0000 (13:21 -0400)]
ext4: Add support for non-native signed/unsigned htree hash algorithms

The original ext3 hash algorithms assumed that variables of type char
were signed, as God and K&R intended.  Unfortunately, this assumption
is not true on some architectures.  Userspace support for marking
filesystems with non-native signed/unsigned chars was added two years
ago, but the kernel-side support was never added (until now).

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
16 years agoext3: Add support for non-native signed/unsigned htree hash algorithms
Theodore Ts'o [Tue, 28 Oct 2008 17:21:55 +0000 (13:21 -0400)]
ext3: Add support for non-native signed/unsigned htree hash algorithms

The original ext3 hash algorithms assumed that variables of type char
were signed, as God and K&R intended.  Unfortunately, this assumption
is not true on some architectures.  Userspace support for marking
filesystems with non-native signed/unsigned chars was added two years
ago, but the kernel-side support was never added (until now).

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: akpm@linux-foundation.org
Cc: linux-kernel@vger.kernel.org
16 years agoext4: fix printk format warning
Alexander Beregalov [Wed, 29 Oct 2008 21:13:08 +0000 (17:13 -0400)]
ext4: fix printk format warning

fs/ext4/balloc.c:607: warning: format '%lld' expects type 'long long int', but argument 2 has type 's64'
fs/ext4/inode.c:1822: warning: format '%lld' expects type 'long long int', but argument 2 has type 's64'
fs/ext4/inode.c:1824: warning: format '%lld' expects type 'long long int', but argument 2 has type 's64'

Signed-off-by: Alexander Beregalov <a.beregalov@gmail.com>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
15 years agoMerge branch 'audit.b61' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/audit...
Linus Torvalds [Mon, 5 Jan 2009 00:32:11 +0000 (16:32 -0800)]
Merge branch 'audit.b61' of git://git./linux/kernel/git/viro/audit-current

* 'audit.b61' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/audit-current:
  audit: validate comparison operations, store them in sane form
  clean up audit_rule_{add,del} a bit
  make sure that filterkey of task,always rules is reported
  audit rules ordering, part 2
  fixing audit rule ordering mess, part 1
  audit_update_lsm_rules() misses the audit_inode_hash[] ones
  sanitize audit_log_capset()
  sanitize audit_fd_pair()
  sanitize audit_mq_open()
  sanitize AUDIT_MQ_SENDRECV
  sanitize audit_mq_notify()
  sanitize audit_mq_getsetattr()
  sanitize audit_ipc_set_perm()
  sanitize audit_ipc_obj()
  sanitize audit_socketcall
  don't reallocate buffer in every audit_sockaddr()

15 years agortc: add alarm/update irq interfaces
Alessandro Zummo [Sun, 4 Jan 2009 20:00:54 +0000 (12:00 -0800)]
rtc: add alarm/update irq interfaces

Add standard interfaces for alarm/update irqs enabling.  Drivers are no
more required to implement equivalent ioctl code as rtc-dev will provide
it.

UIE emulation should now be handled correctly and will work even for those
RTC drivers who cannot be configured to do both UIE and AIE.

Signed-off-by: Alessandro Zummo <a.zummo@towertech.it>
Cc: David Brownell <david-b@pacbell.net>
Cc: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Cc: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agofs: symlink write_begin allocation context fix
Nick Piggin [Sun, 4 Jan 2009 20:00:53 +0000 (12:00 -0800)]
fs: symlink write_begin allocation context fix

With the write_begin/write_end aops, page_symlink was broken because it
could no longer pass a GFP_NOFS type mask into the point where the
allocations happened.  They are done in write_begin, which would always
assume that the filesystem can be entered from reclaim.  This bug could
cause filesystem deadlocks.

The funny thing with having a gfp_t mask there is that it doesn't really
allow the caller to arbitrarily tinker with the context in which it can be
called.  It couldn't ever be GFP_ATOMIC, for example, because it needs to
take the page lock.  The only thing any callers care about is __GFP_FS
anyway, so turn that into a single flag.

Add a new flag for write_begin, AOP_FLAG_NOFS.  Filesystems can now act on
this flag in their write_begin function.  Change __grab_cache_page to
accept a nofs argument as well, to honour that flag (while we're there,
change the name to grab_cache_page_write_begin which is more instructive
and does away with random leading underscores).

This is really a more flexible way to go in the end anyway -- if a
filesystem happens to want any extra allocations aside from the pagecache
ones in ints write_begin function, it may now use GFP_KERNEL (rather than
GFP_NOFS) for common case allocations (eg.  ocfs2_alloc_write_ctxt, for a
random example).

[kosaki.motohiro@jp.fujitsu.com: fix ubifs]
[kosaki.motohiro@jp.fujitsu.com: fix fuse]
Signed-off-by: Nick Piggin <npiggin@suse.de>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: <stable@kernel.org> [2.6.28.x]
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
[ Cleaned up the calling convention: just pass in the AOP flags
  untouched to the grab_cache_page_write_begin() function.  That
  just simplifies everybody, and may even allow future expansion of the
  logic.   - Linus ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoviafb: fix crashes due to 4k stack overflow
Bruno Prémont [Sun, 4 Jan 2009 21:11:54 +0000 (13:11 -0800)]
viafb: fix crashes due to 4k stack overflow

The function viafb_cursor() uses 2 stack-variables of CURSOR_SIZE bits;
CURSOR_SIZE is defined as (8 * 1024).  Using up twice 1k on stack is too
much for 4k-stack (though it works with 8k-stacks).  Make those two
variables kzalloc'ed to preserve stack space.

Also merge the whole lot of local struct's in viafb_ioctl into a union so
the stack usage gets minimized here as well.  (struct's are only accessed
in their indicidual IOCTL case) This second part is only compile-tested as
I know of no userspace app using the IOCTLs.

Signed-off-by: Bruno Prémont <bonbons@linux-vserver.org>
Cc: <JosephChan@via.com.tw>
Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agofs: introduce bgl_lock_ptr()
Pekka Enberg [Sun, 4 Jan 2009 20:00:48 +0000 (12:00 -0800)]
fs: introduce bgl_lock_ptr()

As suggested by Andreas Dilger, introduce a bgl_lock_ptr() helper in
<linux/blockgroup_lock.h> and add separate sb_bgl_lock() helpers to
filesystem specific header files to break the hidden dependency to
struct ext[234]_sb_info.

Also, while at it, convert the macros to static inlines to try make up
for all the times I broke Andrew Morton's tree.

Acked-by: Andreas Dilger <adilger@sun.com>
Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agospi.h uses/needs device.h
Randy Dunlap [Sun, 4 Jan 2009 20:00:47 +0000 (12:00 -0800)]
spi.h uses/needs device.h

Include header files as used/needed:

  In file included from drivers/leds/leds-dac124s085.c:16:
  include/linux/spi/spi.h:66: error: field 'dev' has incomplete type
  include/linux/spi/spi.h: In function 'to_spi_device':
  include/linux/spi/spi.h:100: warning: type defaults to 'int' in declaration of '__mptr'
  ...

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agovmalloc.c: fix flushing in vmap_page_range()
Adam Lackorzynski [Sun, 4 Jan 2009 20:00:46 +0000 (12:00 -0800)]
vmalloc.c: fix flushing in vmap_page_range()

The flush_cache_vmap in vmap_page_range() is called with the end of the
range twice.  The following patch fixes this for me.

Signed-off-by: Adam Lackorzynski <adam@os.inf.tu-dresden.de>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agocgroups: fix a race between cgroup_clone and umount
Li Zefan [Sun, 4 Jan 2009 20:00:45 +0000 (12:00 -0800)]
cgroups: fix a race between cgroup_clone and umount

The race is calling cgroup_clone() while umounting the ns cgroup subsys,
and thus cgroup_clone() might access invalid cgroup_fs, or kill_sb() is
called after cgroup_clone() created a new dir in it.

The BUG I triggered is BUG_ON(root->number_of_cgroups != 1);

  ------------[ cut here ]------------
  kernel BUG at kernel/cgroup.c:1093!
  invalid opcode: 0000 [#1] SMP
  ...
  Process umount (pid: 5177, ti=e411e000 task=e40c4670 task.ti=e411e000)
  ...
  Call Trace:
   [<c0493df7>] ? deactivate_super+0x3f/0x51
   [<c04a3600>] ? mntput_no_expire+0xb3/0xdd
   [<c04a3ab2>] ? sys_umount+0x265/0x2ac
   [<c04a3b06>] ? sys_oldumount+0xd/0xf
   [<c0403911>] ? sysenter_do_call+0x12/0x31
  ...
  EIP: [<c0456e76>] cgroup_kill_sb+0x23/0xe0 SS:ESP 0068:e411ef2c
  ---[ end trace c766c1be3bf944ac ]---

Cc: Serge E. Hallyn <serue@us.ibm.com>
Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Cc: Paul Menage <menage@google.com>
Cc: "Serge E. Hallyn" <serue@us.ibm.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoaudit: validate comparison operations, store them in sane form
Al Viro [Tue, 16 Dec 2008 10:59:26 +0000 (05:59 -0500)]
audit: validate comparison operations, store them in sane form

Don't store the field->op in the messy (and very inconvenient for e.g.
audit_comparator()) form; translate to dense set of values and do full
validation of userland-submitted value while we are at it.

->audit_init_rule() and ->audit_match_rule() get new values now; in-tree
instances updated.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agoclean up audit_rule_{add,del} a bit
Al Viro [Mon, 15 Dec 2008 06:50:28 +0000 (01:50 -0500)]
clean up audit_rule_{add,del} a bit

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agomake sure that filterkey of task,always rules is reported
Al Viro [Tue, 16 Dec 2008 08:51:22 +0000 (03:51 -0500)]
make sure that filterkey of task,always rules is reported

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agoaudit rules ordering, part 2
Al Viro [Mon, 15 Dec 2008 06:17:50 +0000 (01:17 -0500)]
audit rules ordering, part 2

Fix the actual rule listing; add per-type lists _not_ used for matching,
with all exit,... sitting on one such list.  Simplifies "do something
for all rules" logics, while we are at it...

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agofixing audit rule ordering mess, part 1
Al Viro [Mon, 15 Dec 2008 04:45:27 +0000 (23:45 -0500)]
fixing audit rule ordering mess, part 1

Problem: ordering between the rules on exit chain is currently lost;
all watch and inode rules are listed after everything else _and_
exit,never on one kind doesn't stop exit,always on another from
being matched.

Solution: assign priorities to rules, keep track of the current
highest-priority matching rule and its result (always/never).

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agoaudit_update_lsm_rules() misses the audit_inode_hash[] ones
Al Viro [Sun, 14 Dec 2008 17:04:02 +0000 (12:04 -0500)]
audit_update_lsm_rules() misses the audit_inode_hash[] ones

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agosanitize audit_log_capset()
Al Viro [Sun, 4 Jan 2009 19:52:57 +0000 (14:52 -0500)]
sanitize audit_log_capset()

* no allocations
* return void
* don't duplicate checked for dummy context

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agosanitize audit_fd_pair()
Al Viro [Sun, 14 Dec 2008 09:57:47 +0000 (04:57 -0500)]
sanitize audit_fd_pair()

* no allocations
* return void

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agosanitize audit_mq_open()
Al Viro [Sun, 14 Dec 2008 09:02:26 +0000 (04:02 -0500)]
sanitize audit_mq_open()

* don't bother with allocations
* don't do double copy_from_user()
* don't duplicate parts of check for audit_dummy_context()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agosanitize AUDIT_MQ_SENDRECV
Al Viro [Sun, 14 Dec 2008 08:46:48 +0000 (03:46 -0500)]
sanitize AUDIT_MQ_SENDRECV

* logging the original value of *msg_prio in mq_timedreceive(2)
  is insane - the argument is write-only (i.e. syscall always
  ignores the original value and only overwrites it).
* merge __audit_mq_timed{send,receive}
* don't do copy_from_user() twice
* don't mess with allocations in auditsc part
* ... and don't bother checking !audit_enabled and !context in there -
  we'd already checked for audit_dummy_context().

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agosanitize audit_mq_notify()
Al Viro [Wed, 10 Dec 2008 12:16:12 +0000 (07:16 -0500)]
sanitize audit_mq_notify()

* don't copy_from_user() twice
* don't bother with allocations
* don't duplicate parts of audit_dummy_context()
* make it return void

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agosanitize audit_mq_getsetattr()
Al Viro [Wed, 10 Dec 2008 11:58:59 +0000 (06:58 -0500)]
sanitize audit_mq_getsetattr()

* get rid of allocations
* make it return void
* don't duplicate parts of audit_dummy_context()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agosanitize audit_ipc_set_perm()
Al Viro [Wed, 10 Dec 2008 08:47:15 +0000 (03:47 -0500)]
sanitize audit_ipc_set_perm()

* get rid of allocations
* make it return void
* simplify callers

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agosanitize audit_ipc_obj()
Al Viro [Wed, 10 Dec 2008 08:40:06 +0000 (03:40 -0500)]
sanitize audit_ipc_obj()

* get rid of allocations
* make it return void
* simplify callers

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agosanitize audit_socketcall
Al Viro [Wed, 10 Dec 2008 08:16:51 +0000 (03:16 -0500)]
sanitize audit_socketcall

* don't bother with allocations
* now that it can't fail, make it return void

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agodon't reallocate buffer in every audit_sockaddr()
Al Viro [Wed, 10 Dec 2008 00:50:34 +0000 (19:50 -0500)]
don't reallocate buffer in every audit_sockaddr()

No need to do that more than once per process lifetime; allocating/freeing
on each sendto/accept/etc. is bloody pointless.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
15 years agoMerge branch 'cpus4096-for-linus-3' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 3 Jan 2009 20:04:39 +0000 (12:04 -0800)]
Merge branch 'cpus4096-for-linus-3' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'cpus4096-for-linus-3' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (77 commits)
  x86: setup_per_cpu_areas() cleanup
  cpumask: fix compile error when CONFIG_NR_CPUS is not defined
  cpumask: use alloc_cpumask_var_node where appropriate
  cpumask: convert shared_cpu_map in acpi_processor* structs to cpumask_var_t
  x86: use cpumask_var_t in acpi/boot.c
  x86: cleanup some remaining usages of NR_CPUS where s/b nr_cpu_ids
  sched: put back some stack hog changes that were undone in kernel/sched.c
  x86: enable cpus display of kernel_max and offlined cpus
  ia64: cpumask fix for is_affinity_mask_valid()
  cpumask: convert RCU implementations, fix
  xtensa: define __fls
  mn10300: define __fls
  m32r: define __fls
  h8300: define __fls
  frv: define __fls
  cris: define __fls
  cpumask: CONFIG_DISABLE_OBSOLETE_CPUMASK_FUNCTIONS
  cpumask: zero extra bits in alloc_cpumask_var_node
  cpumask: replace for_each_cpu_mask_nr with for_each_cpu in kernel/time/
  cpumask: convert mm/
  ...

15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/linux...
Linus Torvalds [Sat, 3 Jan 2009 20:03:52 +0000 (12:03 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/joro/linux-2.6-iommu

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/linux-2.6-iommu: (89 commits)
  AMD IOMMU: remove now unnecessary #ifdefs
  AMD IOMMU: prealloc_protection_domains should be static
  kvm/iommu: fix compile warning
  AMD IOMMU: add statistics about total number of map requests
  AMD IOMMU: add statistics about allocated io memory
  AMD IOMMU: add stats counter for domain tlb flushes
  AMD IOMMU: add stats counter for single iommu domain tlb flushes
  AMD IOMMU: add stats counter for cross-page request
  AMD IOMMU: add stats counter for free_coherent requests
  AMD IOMMU: add stats counter for alloc_coherent requests
  AMD IOMMU: add stats counter for unmap_sg requests
  AMD IOMMU: add stats counter for map_sg requests
  AMD IOMMU: add stats counter for unmap_single requests
  AMD IOMMU: add stats counter for map_single requests
  AMD IOMMU: add stats counter for completion wait events
  AMD IOMMU: add init code for statistic collection
  AMD IOMMU: add necessary header defines for stats counting
  AMD IOMMU: add Kconfig entry for statistic collection code
  AMD IOMMU: use dev_name in iommu_enable function
  AMD IOMMU: use calc_devid in prealloc_protection_domains
  ...

15 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab...
Linus Torvalds [Sat, 3 Jan 2009 20:02:18 +0000 (12:02 -0800)]
Merge branch 'for_linus' of git://git./linux/kernel/git/mchehab/linux-2.6

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-2.6: (34 commits)
  V4L/DVB (10173): Missing v4l2_prio_close in radio_release
  V4L/DVB (10172): add DVB_DEVICE_TYPE= to uevent
  V4L/DVB (10171): Use usb_set_intfdata
  V4L/DVB (10170): tuner-simple: prevent possible OOPS caused by divide by zero error
  V4L/DVB (10168): sms1xxx: fix inverted gpio for lna control on tiger r2
  V4L/DVB (10167): sms1xxx: add support for inverted gpio
  V4L/DVB (10166): dvb frontend: stop using non-C99 compliant comments
  V4L/DVB (10165): Add FE_CAN_2G_MODULATION flag to frontends that support DVB-S2
  V4L/DVB (10164): Add missing S2 caps flag to S2API
  V4L/DVB (10163): em28xx: allocate adev together with struct em28xx dev
  V4L/DVB (10162): tuner-simple: Fix tuner type set message
  V4L/DVB (10161): saa7134: fix autodetection for AVer TV GO 007 FM Plus
  V4L/DVB (10160): em28xx: update chip id for em2710
  V4L/DVB (10157): Add USB ID for the Sil4701 radio from DealExtreme
  V4L/DVB (10156): saa7134: Add support for Avermedia AVer TV GO 007 FM Plus
  V4L/DVB (10155): Add TEA5764 radio driver
  V4L/DVB (10154): saa7134: fix a merge conflict on Behold H6 board
  V4L/DVB (10153): Add the Beholder H6 card to DVB-T part of sources.
  V4L/DVB (10152): Change configuration of the Beholder H6 card
  V4L/DVB (10151): Fix I2C bridge error in zl10353
  ...

15 years agosparseirq: move set/get_timer_rand_state back to .c
Yinghai Lu [Sat, 3 Jan 2009 08:06:34 +0000 (00:06 -0800)]
sparseirq: move set/get_timer_rand_state back to .c

those two functions only used in that C file

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/drzeus/mmc
Linus Torvalds [Sat, 3 Jan 2009 20:00:07 +0000 (12:00 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/drzeus/mmc

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/drzeus/mmc:
  mmc: warn about voltage mismatches
  mmc_spi: Add support for OpenFirmware bindings
  pxamci: fix dma_unmap_sg length
  mmc_block: ensure all sectors that do not have errors are read
  drivers/mmc: Move a dereference below a NULL test
  sdhci: handle built-in sdhci with modular leds class
  mmc: balanc pci_iomap with pci_iounmap
  mmc_block: print better error messages
  mmc: Add mmc_vddrange_to_ocrmask() helper function
  ricoh_mmc: Handle newer models of Ricoh controllers
  mmc: Add 8-bit bus width support
  sdhci: activate led support also when module
  mmc: trivial annotation of 'blocks'
  pci: use pci_ioremap_bar() in drivers/mmc
  sdricoh_cs: Add support for Bay Controller devices
  mmc: at91_mci: reorder timer setup and mmc_add_host() call

15 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/wim/linux-2.6-watchdog
Linus Torvalds [Sat, 3 Jan 2009 19:59:13 +0000 (11:59 -0800)]
Merge git://git./linux/kernel/git/wim/linux-2.6-watchdog

* git://git.kernel.org/pub/scm/linux/kernel/git/wim/linux-2.6-watchdog:
  [WATCHDOG] Add support for the WM8350 watchdog
  [WATCHDOG] Add SMSC SCH311x Watchdog Timer.
  [WATCHDOG] ib700wdt - add timeout parameter

15 years agoMerge branch 'cputime' of git://git390.osdl.marist.edu/pub/scm/linux-2.6
Linus Torvalds [Sat, 3 Jan 2009 19:56:24 +0000 (11:56 -0800)]
Merge branch 'cputime' of git://git390.osdl.marist.edu/linux-2.6

* 'cputime' of git://git390.osdl.marist.edu/pub/scm/linux-2.6:
  [PATCH] fast vdso implementation for CLOCK_THREAD_CPUTIME_ID
  [PATCH] improve idle cputime accounting
  [PATCH] improve precision of idle time detection.
  [PATCH] improve precision of process accounting.
  [PATCH] idle cputime accounting
  [PATCH] fix scaled & unscaled cputime accounting

15 years agoMake %p print '(null)' for NULL pointers
Linus Torvalds [Sat, 3 Jan 2009 19:46:17 +0000 (11:46 -0800)]
Make %p print '(null)' for NULL pointers

Before, when we only ever printed out the pointer value itself, a NULL
pointer would never cause issues and might as well be printed out as
just its numeric value.

However, with the extended %p formats, especially %pR, we might validly
want to print out resources for debugging.  And sometimes they don't
even exist, and the resource pointer is just NULL.  Print it out as
such, rather than oopsing.

This is a more generic version of a patch done by Trent Piepho (catching
all %p cases rather than just %pR, and using "(null)" instead of
"[NULL]" to match glibc).

Requested-by: Trent Piepho <xyzzy@speakeasy.org>
Acked-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoget rid of special-casing the /sbin/loader on alpha
Al Viro [Sat, 3 Jan 2009 07:16:33 +0000 (07:16 +0000)]
get rid of special-casing the /sbin/loader on alpha

... just make it a binfmt handler like #! one.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agosanitize ifdefs in binfmt_aout
Al Viro [Sat, 3 Jan 2009 07:16:23 +0000 (07:16 +0000)]
sanitize ifdefs in binfmt_aout

They are actually alpha vs.  i386/arm/m68k i.e. ecoff vs. aout.

In the only place where we actually tried to handle arm and i386/m68k in
different ways (START_DATA() in coredump handling), the arm variant
works for all of them (i386 and m68k have u.start_code set to 0).

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoremove the rudiment of a.out for sparc
Al Viro [Sat, 3 Jan 2009 07:16:13 +0000 (07:16 +0000)]
remove the rudiment of a.out for sparc

it's been used only in sunos compat

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agoUpdated contact info for CREDITS file
Hugh Blemings [Sat, 3 Jan 2009 05:48:44 +0000 (16:48 +1100)]
Updated contact info for CREDITS file

This updates some personal info in the CREDITS file.

I'm no longer actively involved in Keyspan driver work so shouldn't
really be listed as a Maintainer here.

I do however field the occasional question on them and as I'm dropping
the misc.nu domain, want to ensure people can find me should they need
to.

Signed-off-by: Hugh Blemings <hugh@blemings.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 years agox86: setup_per_cpu_areas() cleanup
Cyrill Gorcunov [Fri, 2 Jan 2009 18:51:32 +0000 (21:51 +0300)]
x86: setup_per_cpu_areas() cleanup

Impact: cleanup

__alloc_bootmem and __alloc_bootmem_node do panic
for us in case of fail so no need for additional
checks here.

Also lets use pr_*() macros for printing.

Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agocpumask: fix compile error when CONFIG_NR_CPUS is not defined
Mike Travis [Thu, 1 Jan 2009 02:08:48 +0000 (18:08 -0800)]
cpumask: fix compile error when CONFIG_NR_CPUS is not defined

CONFIG_NR_CPUS will be defined for all arch's whether SMP or not, but
it may not have made it into all arches yet.

Signed-off-by: Mike Travis <travis@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agocpumask: use alloc_cpumask_var_node where appropriate
Mike Travis [Thu, 1 Jan 2009 02:08:47 +0000 (18:08 -0800)]
cpumask: use alloc_cpumask_var_node where appropriate

Impact: Reduce inter-node memory traffic.

Reduces inter-node memory traffic (offloading the global system bus)
by allocating referenced struct cpumasks on the same node as the
referring struct.

Signed-off-by: Mike Travis <travis@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agocpumask: convert shared_cpu_map in acpi_processor* structs to cpumask_var_t
Rusty Russell [Thu, 1 Jan 2009 02:08:47 +0000 (18:08 -0800)]
cpumask: convert shared_cpu_map in acpi_processor* structs to cpumask_var_t

Impact: Reduce memory usage, use new API.

This is part of an effort to reduce structure sizes for machines
configured with large NR_CPUS.  cpumask_t gets replaced by
cpumask_var_t, which is either struct cpumask[1] (small NR_CPUS) or
struct cpumask * (large NR_CPUS).

(Changes to powernow-k* by <travis>.)

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Mike Travis <travis@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: use cpumask_var_t in acpi/boot.c
Rusty Russell [Thu, 1 Jan 2009 02:08:47 +0000 (18:08 -0800)]
x86: use cpumask_var_t in acpi/boot.c

Impact: reduce stack size, use new API.

Replace cpumask_t with cpumask_var_t.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Mike Travis <travis@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agox86: cleanup some remaining usages of NR_CPUS where s/b nr_cpu_ids
Mike Travis [Thu, 1 Jan 2009 02:08:46 +0000 (18:08 -0800)]
x86: cleanup some remaining usages of NR_CPUS where s/b nr_cpu_ids

Impact: Reduce future system panics due to cpumask operations using NR_CPUS

Insure that code does not look at bits >= nr_cpu_ids as when cpumasks are
allocated based on nr_cpu_ids, these extra bits will not be defined.

Also some other minor updates:

   * change in to use cpu accessor function set_cpu_present() instead of
     directly accessing cpu_present_map w/cpu_clear() [arch/x86/kernel/reboot.c]

   * use cpumask_of() instead of &cpumask_of_cpu() [arch/x86/kernel/reboot.c]

   * optimize some cpu_mask_to_apicid_and functions.

Signed-off-by: Mike Travis <travis@sgi.com>
Acked-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
15 years agosched: put back some stack hog changes that were undone in kernel/sched.c
Mike Travis [Thu, 1 Jan 2009 02:08:45 +0000 (18:08 -0800)]
sched: put back some stack hog changes that were undone in kernel/sched.c

Impact: prevents panic from stack overflow on numa-capable machines.

Some of the "removal of stack hogs" changes in kernel/sched.c by using
node_to_cpumask_ptr were undone by the early cpumask API updates, and
causes a panic due to stack overflow.  This patch undoes those changes
by using cpumask_of_node() which returns a 'const struct cpumask *'.

In addition, cpu_coregoup_map is replaced with cpu_coregroup_mask further
reducing stack usage.  (Both of these updates removed 9 FIXME's!)

Also:
   Pick up some remaining changes from the old 'cpumask_t' functions to
   the new 'struct cpumask *' functions.

   Optimize memory traffic by allocating each percpu local_cpu_mask on the
   same node as the referring cpu.

Signed-off-by: Mike Travis <travis@sgi.com>
Acked-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Ingo Molnar <mingo@elte.hu>