firefly-linux-kernel-4.4.55.git
16 years ago[XFS] kill deleted inodes list
David Chinner [Thu, 30 Oct 2008 06:37:49 +0000 (17:37 +1100)]
[XFS] kill deleted inodes list

Now that the deleted inodes list is unused, kill it. This also removes the
i_reclaim list head from the xfs_inode, shrinking it by two pointers.

SGI-PV: 988142

SGI-Modid: xfs-linux-melb:xfs-kern:32334a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] use the inode radix tree for reclaiming inodes
David Chinner [Thu, 30 Oct 2008 06:37:37 +0000 (17:37 +1100)]
[XFS] use the inode radix tree for reclaiming inodes

Use the reclaim tag to walk the radix tree and find the inodes under
reclaim. This was the only user of the deleted inode list.

SGI-PV: 988142

SGI-Modid: xfs-linux-melb:xfs-kern:32333a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] mark inodes for reclaim via a tag in the inode radix tree
David Chinner [Thu, 30 Oct 2008 06:37:26 +0000 (17:37 +1100)]
[XFS] mark inodes for reclaim via a tag in the inode radix tree

Prepare for removing the deleted inode list by marking inodes for reclaim
in the inode radix trees so that we can use the radix trees to find
reclaimable inodes.

SGI-PV: 988142

SGI-Modid: xfs-linux-melb:xfs-kern:32331a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] rename inode reclaim functions
David Chinner [Thu, 30 Oct 2008 06:37:15 +0000 (17:37 +1100)]
[XFS] rename inode reclaim functions

The function names xfs_finish_reclaim and xfs_finish_reclaim_all are not
very descriptive of what they are reclaiming. Rename to
xfs_reclaim_inode[s] to match the xfs_sync_inodes() function.

SGI-PV: 988142

SGI-Modid: xfs-linux-melb:xfs-kern:32330a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] move inode reclaim functions to xfs_sync.c
David Chinner [Thu, 30 Oct 2008 06:37:03 +0000 (17:37 +1100)]
[XFS] move inode reclaim functions to xfs_sync.c

Background inode reclaim is run by the xfssyncd. Move the reclaim worker
functions to be close to the sync code as the are very similar in
structure and are both run from the same background thread.

SGI-PV: 988142

SGI-Modid: xfs-linux-melb:xfs-kern:32329a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] Fix build warning - xfs_fs_alloc_inode() needs a return statement
Lachlan McIlroy [Thu, 30 Oct 2008 06:36:52 +0000 (17:36 +1100)]
[XFS] Fix build warning - xfs_fs_alloc_inode() needs a return statement

SGI-PV: 988141

SGI-Modid: xfs-linux-melb:xfs-kern:32325a

Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Prevent use-after-free caused by synchronous inode reclaim
David Chinner [Thu, 30 Oct 2008 06:36:40 +0000 (17:36 +1100)]
[XFS] Prevent use-after-free caused by synchronous inode reclaim

With the combined linux and XFS inode, we need to ensure that the combined
structure is not freed before the generic code is finished with the inode.
As it turns out, there is a case where the XFS inode is freed before the
linux inode - when xfs_reclaim() is called from ->clear_inode() on a clean
inode, the xfs inode is freed during that call. The generic code
references the inode after the ->clear_inode() call, so this is a use
after free situation.

Fix the problem by moving the xfs_reclaim() call to ->destroy_inode()
instead of in ->clear_inode(). This ensures the combined inode structure
is not freed until after the generic code has finished with it.

SGI-PV: 988141

SGI-Modid: xfs-linux-melb:xfs-kern:32324a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] Combine the XFS and Linux inodes
David Chinner [Thu, 30 Oct 2008 06:36:14 +0000 (17:36 +1100)]
[XFS] Combine the XFS and Linux inodes

To avoid issues with different lifecycles of XFS and Linux inodes, embedd
the linux inode inside the XFS inode. This means that the linux inode has
the same lifecycle as the XFS inode, even when it has been released by the
OS. XFS inodes don't live much longer than this (a short stint in reclaim
at most), so there isn't significant memory usage penalties here.

Version 3 o kill xfs_icount()

Version 2 o remove unused commented out code from xfs_iget(). o kill
useless cast in VFS_I()

SGI-PV: 988141

SGI-Modid: xfs-linux-melb:xfs-kern:32323a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years agoInode: Allow external list initialisation
David Chinner [Thu, 30 Oct 2008 06:35:24 +0000 (17:35 +1100)]
Inode: Allow external list initialisation

To allow XFS to combine the XFS and linux inodes into a single
structure, we need to drive inode lookup from the XFS inode cache,
not the generic inode cache. This means that we need initialise a
struct inode from a context outside alloc_inode() as it is no longer
used by XFS.

After inode allocation and initialisation, we need to add the inode
to the superblock list, the in-use list, hash it and do some
accounting. This all needs to be done with the inode_lock held and
there are already several places in fs/inode.c that do this list
manipulation.  Factor out the common code, add a locking wrapper and
export the function so ti can be called from XFS.

Signed-off-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years agoInode: Allow external initialisers
David Chinner [Thu, 30 Oct 2008 06:32:23 +0000 (17:32 +1100)]
Inode: Allow external initialisers

To allow XFS to combine the XFS and linux inodes into a single
structure, we need to drive inode lookup from the XFS inode cache,
not the generic inode cache. This means that we need initialise a
struct inode from a context outside alloc_inode() as it is no longer
used by XFS.

Factor and export the struct inode initialisation code from
alloc_inode() to inode_init_always() as a counterpart to
inode_init_once().  i.e. we have to call this init function for each
inode instantiation (always), as opposed inode_init_once() which is
only called on slab object instantiation (once).

Signed-off-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Never call mark_inode_dirty_sync() directly
David Chinner [Thu, 30 Oct 2008 06:21:30 +0000 (17:21 +1100)]
[XFS] Never call mark_inode_dirty_sync() directly

Once the Linux inode and the XFS inode are combined, we cannot rely on
just check if the linux inode exists as a method of determining if it is
valid or not. Hence we should always call xfs_mark_inode_dirty_sync()
instead as it does the correct checks to determine if the liinux inode is
in a valid state or not.

SGI-PV: 988141

SGI-Modid: xfs-linux-melb:xfs-kern:32318a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] factor xfs_iget_core() into hit and miss cases
David Chinner [Thu, 30 Oct 2008 06:21:19 +0000 (17:21 +1100)]
[XFS] factor xfs_iget_core() into hit and miss cases

There are really two cases in xfs_iget_core(). The first is the cache hit
case, the second is the miss case. They share very little code, and hence
can easily be factored out into separate functions. This makes the code
much easier to understand and subsequently modify.

SGI-PV: 988141

SGI-Modid: xfs-linux-melb:xfs-kern:32317a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] fix instant oops with tracing enabled
Christoph Hellwig [Thu, 30 Oct 2008 06:21:10 +0000 (17:21 +1100)]
[XFS] fix instant oops with tracing enabled

We can only read inode->i_count if the inode is actually there and not a
NULL pointer. This was introduced in one of the recent sync patches.

SGI-PV: 988255

SGI-Modid: xfs-linux-melb:xfs-kern:32315a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Move remaining quiesce code.
David Chinner [Thu, 30 Oct 2008 06:16:21 +0000 (17:16 +1100)]
[XFS] Move remaining quiesce code.

With all the other filesystem sync code it in xfs_sync.c including the
data quiesce code, it makes sense to move the remaining quiesce code to
the same place.

SGI-PV: 988140

SGI-Modid: xfs-linux-melb:xfs-kern:32312a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] Kill xfs_sync()
David Chinner [Thu, 30 Oct 2008 06:16:11 +0000 (17:16 +1100)]
[XFS] Kill xfs_sync()

There are no more callers to xfs_sync() now, so remove the function
altogther.

SGI-PV: 988140

SGI-Modid: xfs-linux-melb:xfs-kern:32311a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] Kill SYNC_CLOSE
David Chinner [Thu, 30 Oct 2008 06:16:00 +0000 (17:16 +1100)]
[XFS] Kill SYNC_CLOSE

SYNC_CLOSE is only ever used and checked in conjunction with SYNC_WAIT,
and this only done in one spot. The only thing this does is make
XFS_bflush() calls to the data buftargs.

This will happen very shortly afterwards the xfs_sync() call anyway in the
unmount path via the xfs_close_devices(), so this code is redundant and
can be removed. That only user of SYNC_CLOSE is now gone, so kill the flag
completely.

SGI-PV: 988140

SGI-Modid: xfs-linux-melb:xfs-kern:32310a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] make SYNC_DELWRI no longer use xfs_sync
David Chinner [Thu, 30 Oct 2008 06:15:50 +0000 (17:15 +1100)]
[XFS] make SYNC_DELWRI no longer use xfs_sync

Continue to de-multiplex xfs_sync be replacing all SYNC_DELWRI callers
with direct calls functions that do the work. Isolate the data quiesce
case to a function in xfs_sync.c. Isolate the FSDATA case with explicit
calls to xfs_sync_fsdata().

Version 2: o Push delwri related log forces into xfs_sync_inodes().

SGI-PV: 988140

SGI-Modid: xfs-linux-melb:xfs-kern:32309a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] make SYNC_ATTR no longer use xfs_sync
David Chinner [Thu, 30 Oct 2008 06:15:38 +0000 (17:15 +1100)]
[XFS] make SYNC_ATTR no longer use xfs_sync

Continue to de-multiplex xfs_sync be replacing all SYNC_ATTR callers with
direct calls xfs_sync_inodes(). Add an assert into xfs_sync() to ensure we
caught all the SYNC_ATTR callers.

SGI-PV: 988140

SGI-Modid: xfs-linux-melb:xfs-kern:32308a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] xfssyncd: don't call xfs_sync
David Chinner [Thu, 30 Oct 2008 06:15:29 +0000 (17:15 +1100)]
[XFS] xfssyncd: don't call xfs_sync

Start de-multiplexing xfs_sync() by making xfs_sync_worker() call the
specific sync functions it needs. This is only a small, unique subset of
the entire xfs_sync() code so is easier to follow.

SGI-PV: 988140

SGI-Modid: xfs-linux-melb:xfs-kern:32307a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] kill xfs_syncsub
David Chinner [Thu, 30 Oct 2008 06:15:21 +0000 (17:15 +1100)]
[XFS] kill xfs_syncsub

Now that the only caller is xfs_sync(), merge the two together as it makes
no sense to keep them separate.

SGI-PV: 988140

SGI-Modid: xfs-linux-melb:xfs-kern:32306a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] use xfs_sync_inodes rather than xfs_syncsub
David Chinner [Thu, 30 Oct 2008 06:15:12 +0000 (17:15 +1100)]
[XFS] use xfs_sync_inodes rather than xfs_syncsub

Kill the unused arg in xfs_syncsub() and xfs_sync_inodes(). For callers of
xfs_syncsub() that only want to flush inodes, replace xfs_syncsub() with
direct calls to xfs_sync_inodes() as that is all that is being done with
the specific flags being passed in.

SGI-PV: 988140

SGI-Modid: xfs-linux-melb:xfs-kern:32305a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] Use struct inodes instead of vnodes to kill vn_grab
David Chinner [Thu, 30 Oct 2008 06:15:03 +0000 (17:15 +1100)]
[XFS] Use struct inodes instead of vnodes to kill vn_grab

With the sync code relocated to the linux-2.6 directory we can use struct
inodes directly. If we do the same thing for the quota release code, we
can remove vn_grab altogether. While here, convert the VN_BAD() checks to
is_bad_inode() so we can remove vnodes entirely from this code.

SGI-PV: 988140

SGI-Modid: xfs-linux-melb:xfs-kern:32304a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] split out two helpers from xfs_syncsub
Christoph Hellwig [Thu, 30 Oct 2008 06:14:53 +0000 (17:14 +1100)]
[XFS] split out two helpers from xfs_syncsub

Split out two helpers from xfs_syncsub for the dummy log commit and the
superblock writeout.

SGI-PV: 988140

SGI-Modid: xfs-linux-melb:xfs-kern:32303a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Move XFS_BMAP_SANITY_CHECK out of line.
Christoph Hellwig [Thu, 30 Oct 2008 06:14:43 +0000 (17:14 +1100)]
[XFS] Move XFS_BMAP_SANITY_CHECK out of line.

Move the XFS_BMAP_SANITY_CHECK macro out of line and make it a properly
typed function. Also pass the xfs_buf for the btree block instead of just
the btree block header, as we will need some additional information for it
to implement CRC checking of btree blocks.

SGI-PV: 988146

SGI-Modid: xfs-linux-melb:xfs-kern:32301a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Donald Douwsma <donaldd@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Always use struct xfs_btree_block instead of short / longform
Christoph Hellwig [Thu, 30 Oct 2008 06:14:34 +0000 (17:14 +1100)]
[XFS] Always use struct xfs_btree_block instead of short / longform
structures.

Always use the generic xfs_btree_block type instead of the short / long
structures. Add XFS_BTREE_SBLOCK_LEN / XFS_BTREE_LBLOCK_LEN defines for
the length of a short / long form block. The rationale for this is that we
will grow more btree block header variants to support CRCs and other RAS
information, and always accessing them through the same datatype with
unions for the short / long form pointers makes implementing this much
easier.

SGI-PV: 988146

SGI-Modid: xfs-linux-melb:xfs-kern:32300a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Donald Douwsma <donaldd@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] cleanup btree record / key / ptr addressing macros.
Christoph Hellwig [Thu, 30 Oct 2008 06:11:40 +0000 (17:11 +1100)]
[XFS] cleanup btree record / key / ptr addressing macros.

Replace the generic record / key / ptr addressing macros that use cpp
token pasting with simpler macros that do the job for just one given btree
type. The new macros lose the cur argument and thus can be used outside
the core btree code, but also gain an xfs_mount * argument to allow for
checking the CRC flag in the near future. Note that many of these macros
aren't actually used in the kernel code, but only in userspace (mostly in
xfs_repair).

SGI-PV: 988146

SGI-Modid: xfs-linux-melb:xfs-kern:32295a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Donald Douwsma <donaldd@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] remove the mount inode list
David Chinner [Thu, 30 Oct 2008 06:11:29 +0000 (17:11 +1100)]
[XFS] remove the mount inode list

Now we've removed all users of the mount inode list, we can kill it. This
reduces the size of the xfs_inode by 2 pointers.

SGI-PV: 988139

SGI-Modid: xfs-linux-melb:xfs-kern:32293a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] Cleanup maxrecs calculation.
Christoph Hellwig [Thu, 30 Oct 2008 06:11:19 +0000 (17:11 +1100)]
[XFS] Cleanup maxrecs calculation.

Clean up the way the maximum and minimum records for the btree blocks are
calculated. For the alloc and inobt btrees all the values are
pre-calculated in xfs_mount_common, and we switch the current loop around
the ugly generic macros that use cpp token pasting to generate type names
to two small helpers in normal C code. For the bmbt and bmdr trees these
helpers also exist, but can be called during runtime, too. Here we also
kill various macros dealing with them and inline the logic into the
get_minrecs / get_maxrecs / get_dmaxrecs methods in xfs_bmap_btree.c.

Note that all these new helpers take an xfs_mount * argument which will be
needed to determine the size of a btree block once we add support for
extended btree blocks with CRCs and other RAS information.

SGI-PV: 988146

SGI-Modid: xfs-linux-melb:xfs-kern:32292a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Donald Douwsma <donaldd@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Traverse inode trees when releasing dquots
David Chinner [Thu, 30 Oct 2008 06:08:03 +0000 (17:08 +1100)]
[XFS] Traverse inode trees when releasing dquots

Make releasing all inode dquots traverse the per-ag inode radix trees
rather than the mount inode list. This removes another user of the mount
inode list.

Version 3 o fix comment relating to avoiding trying to release the

quota inodes and those in reclaim.

Version 2 o add comment explaining use of gang lookups for a single inode
o use IRELE, not VN_RELE o move check for ag initialisation to caller.

SGI-PV: 988139

SGI-Modid: xfs-linux-melb:xfs-kern:32291a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] Use the inode tree for finding dirty inodes
David Chinner [Thu, 30 Oct 2008 06:07:29 +0000 (17:07 +1100)]
[XFS] Use the inode tree for finding dirty inodes

Update xfs_sync_inodes to walk the inode radix tree cache to find dirty
inodes. This removes a huge bunch of nasty, messy code for traversing the
mount inode list safely and removes another user of the mount inode list.

Version 3 o rediff against new linux-2.6/xfs_sync.c code

Version 2 o add comment explaining use of gang lookups for a single inode
o use IRELE, not VN_RELE o move check for ag initialisation to caller.

SGI-PV: 988139

SGI-Modid: xfs-linux-melb:xfs-kern:32290a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] don't block in xfs_qm_dqflush() during async writeback.
David Chinner [Thu, 30 Oct 2008 06:07:20 +0000 (17:07 +1100)]
[XFS] don't block in xfs_qm_dqflush() during async writeback.

Normally dquots are written back via delayed write mechanisms. They are
flushed to their backing buffer by xfssyncd, which is then pushed out by
either AIL or xfsbufd flushing. The flush from the xfssyncd is supposed to
be non-blocking, but xfs_qm_dqflush() always waits for pinned duots, which
means that it will block for the length of time it takes to do a
synchronous log force. This causes unnecessary extra log I/O to be issued
whenever we try to flush a busy dquot.

Avoid the log forces and blocking xfssyncd by making xfs_qm_dqflush() pay
attention to what type of sync it is doing when it sees a pinned dquot and
not waiting when doing non-blocking flushes.

SGI-PV: 988147

SGI-Modid: xfs-linux-melb:xfs-kern:32287a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Peter Leckie <pleckie@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Remove xfs_iflush_all and clean up xfs_finish_reclaim_all()
David Chinner [Thu, 30 Oct 2008 06:06:28 +0000 (17:06 +1100)]
[XFS] Remove xfs_iflush_all and clean up xfs_finish_reclaim_all()

xfs_iflush_all() walks the m_inodes list to find inodes that need
reclaiming. We already have such a list - the m_del_inodes list. Replace
xfs_iflush_all() with a call to xfs_finish_reclaim_all() and clean up
xfs_finish_reclaim_all() to handle the different flush modes now needed.

Originally based on a patch from Christoph Hellwig.

Version 3 o rediff against new linux-2.6/xfs_sync.c code

Version 2 o revert xfs_syncsub() inode reclaim behaviour back to original

code o xfs_quiesce_fs() should use XFS_IFLUSH_DELWRI_ELSE_ASYNC, not

XFS_IFLUSH_ASYNC, to prevent change of behaviour.

SGI-PV: 988139

SGI-Modid: xfs-linux-melb:xfs-kern:32284a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] move xfssyncd code to xfs_sync.c
David Chinner [Thu, 30 Oct 2008 06:06:18 +0000 (17:06 +1100)]
[XFS] move xfssyncd code to xfs_sync.c

Move all the xfssyncd code to the new xfs_sync.c file. This places it
closer to the actual code that it interacts with, rather than just being
associated with high level VFS code.

SGI-PV: 988139

SGI-Modid: xfs-linux-melb:xfs-kern:32283a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] move sync code to its own file
David Chinner [Thu, 30 Oct 2008 06:06:08 +0000 (17:06 +1100)]
[XFS] move sync code to its own file

The sync code in XFS is spread around several files. While it used to make
sense to have such a distribution, the code is about to be cleaned up and
so centralising it in one spot as the first step makes sense.

SGI-PV: 988139

SGI-Modid: xfs-linux-melb:xfs-kern:32282a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] Show buffer address with debug hexdump on corruption
Barry Naujok [Thu, 30 Oct 2008 06:05:58 +0000 (17:05 +1100)]
[XFS] Show buffer address with debug hexdump on corruption

SGI-PV: 987246

SGI-Modid: xfs-linux-melb:xfs-kern:32233a

Signed-off-by: Barry Naujok <bnaujok@sgi.com>
Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Check agf_btreeblks is valid when reading in the AGF
Barry Naujok [Thu, 30 Oct 2008 06:05:49 +0000 (17:05 +1100)]
[XFS] Check agf_btreeblks is valid when reading in the AGF

SGI-PV: 987683

SGI-Modid: xfs-linux-melb:xfs-kern:32232a

Signed-off-by: Barry Naujok <bnaujok@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Sync up kernel and user-space headers
Barry Naujok [Thu, 30 Oct 2008 06:05:38 +0000 (17:05 +1100)]
[XFS] Sync up kernel and user-space headers

SGI-PV: 986558

SGI-Modid: xfs-linux-melb:xfs-kern:32231a

Signed-off-by: Barry Naujok <bnaujok@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Make xfs_btree_check_ptr() debug-only code.
Lachlan McIlroy [Thu, 30 Oct 2008 06:05:26 +0000 (17:05 +1100)]
[XFS] Make xfs_btree_check_ptr() debug-only code.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32224a

Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] Fix build brakage from patch "Clean up dquot pincount code"
Peter Leckie [Thu, 30 Oct 2008 06:05:18 +0000 (17:05 +1100)]
[XFS] Fix build brakage from patch "Clean up dquot pincount code"

This is a fix for patch " Clean up dquot pincount code" which introduced a
build breakage due to a missing & in xfs_qm_dquot_logitem_pin.

SGI-PV: 986789

SGI-Modid: xfs-linux-melb:xfs-kern:32221a

Signed-off-by: Peter Leckie <pleckie@sgi.com>
Signed-off-by: Donald Douwsma <donaldd@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Clean up dquot pincount code.
Peter Leckie [Thu, 30 Oct 2008 06:05:04 +0000 (17:05 +1100)]
[XFS] Clean up dquot pincount code.

This is a code cleanup and optimization that removes a per mount point
spinlock from the quota code and cleans up the code.

The patch changes the pincount from being an int protected by a spinlock
to an atomic_t allowing the pincount to be manipulated without holding the
spinlock.

This cleanup also protects against random wakup's of both the aild and
xfssyncd by reevaluating the pincount after been woken. Two latter patches
will address the Spurious wakeups.

SGI-PV: 986789

SGI-Modid: xfs-linux-melb:xfs-kern:32215a

Signed-off-by: Peter Leckie <pleckie@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Donald Douwsma <donaldd@sgi.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Wait for all I/O on truncate to zero file size
Lachlan McIlroy [Thu, 30 Oct 2008 05:59:06 +0000 (16:59 +1100)]
[XFS] Wait for all I/O on truncate to zero file size

It's possible to have outstanding xfs_ioend_t's queued when the file size
is zero. This can happen in the direct I/O path when a direct I/O write
fails due to ENOSPC. In this case the xfs_ioend_t will still be queued (ie
xfs_end_io_direct() does not know that the I/O failed so can't force the
xfs_ioend_t to be flushed synchronously).

When we truncate a file on unlink we don't know to wait for these
xfs_ioend_ts and we can have a use-after-free situation if the inode is
reclaimed before the xfs_ioend_t is finally processed.

As was suggested by Dave Chinner lets wait for all I/Os to complete when
truncating the file size to zero.

SGI-PV: 981668

SGI-Modid: xfs-linux-melb:xfs-kern:32216a

Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] make btree tracing generic
Christoph Hellwig [Thu, 30 Oct 2008 05:58:50 +0000 (16:58 +1100)]
[XFS] make btree tracing generic

Make the existing bmap btree tracing generic so that it applies to all
btree types.

Some fragments lifted from a patch by Dave Chinner.

This adds two files that were missed from the previous btree tracing
checkin.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32210a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] mark various functions in xfs_btree.c static
Christoph Hellwig [Thu, 30 Oct 2008 05:58:41 +0000 (16:58 +1100)]
[XFS] mark various functions in xfs_btree.c static

Lots of functionality in xfs_btree.c isn't needed by callers outside of
this file anymore, so mark these functions static.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32209a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] add keys_inorder and recs_inorder btree methods
Christoph Hellwig [Thu, 30 Oct 2008 05:58:32 +0000 (16:58 +1100)]
[XFS] add keys_inorder and recs_inorder btree methods

Add methods to check whether two keys/records are in the righ order. This
replaces the xfs_btree_check_key and xfs_btree_check_rec methods. For the
callers from xfs_bmap.c just opencode the bmbt-specific asserts.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32208a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] kill xfs_bmbt_log_block and xfs_bmbt_log_recs
Christoph Hellwig [Thu, 30 Oct 2008 05:58:21 +0000 (16:58 +1100)]
[XFS] kill xfs_bmbt_log_block and xfs_bmbt_log_recs

These are equivalent to the xfs_btree_* versions, and the only remaining
caller can be switched to the generic one after they are exported. Also
remove some now dead infrastructure in xfs_bmap_btree.c.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32207a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] implement generic xfs_btree_get_rec
Christoph Hellwig [Thu, 30 Oct 2008 05:58:11 +0000 (16:58 +1100)]
[XFS] implement generic xfs_btree_get_rec

Not really much reason to make it generic given that it's so small, but
this is the last non-method in xfs_alloc_btree.c and xfs_ialloc_btree.c,
so it makes the whole btree implementation more structured.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32206a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] implement generic xfs_btree_delete/delrec
Christoph Hellwig [Thu, 30 Oct 2008 05:58:01 +0000 (16:58 +1100)]
[XFS] implement generic xfs_btree_delete/delrec

Make the btree delete code generic. Based on a patch from David Chinner
with lots of changes to follow the original btree implementations more
closely. While this loses some of the generic helper routines for
inserting/moving/removing records it also solves some of the one off bugs
in the original code and makes it easier to verify.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32205a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] move xfs_bmbt_killroot to common code
Christoph Hellwig [Thu, 30 Oct 2008 05:57:51 +0000 (16:57 +1100)]
[XFS] move xfs_bmbt_killroot to common code

xfs_bmbt_killroot is a mostly generic implementation of moving from a real
block based root to an inode based root. So move it to xfs_btree.c where
it can use all the nice infrastructure there and make it pointer size
agnostic

The new name for it is xfs_btree_kill_iroot, following the old naming but
making it clear we're dealing with the root in inode case here, and to
avoid confusion with xfs_btree_new_root which is used for the not inode
rooted case. I've also added a comment describing what it does and why
it's named the way it is.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32203a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] implement generic xfs_btree_insert/insrec
Christoph Hellwig [Thu, 30 Oct 2008 05:57:40 +0000 (16:57 +1100)]
[XFS] implement generic xfs_btree_insert/insrec

Make the btree insert code generic. Based on a patch from David Chinner
with lots of changes to follow the original btree implementations more
closely. While this loses some of the generic helper routines for
inserting/moving/removing records it also solves some of the one off bugs
in the original code and makes it easier to verify.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32202a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] move xfs_bmbt_newroot to common code
Christoph Hellwig [Thu, 30 Oct 2008 05:57:28 +0000 (16:57 +1100)]
[XFS] move xfs_bmbt_newroot to common code

xfs_bmbt_newroot is a mostly generic implementation of moving from an
inode root to a real block based root. So move it to xfs_btree.c where it
can use all the nice infrastructure there and make it pointer size
agnostic

The new name for it is xfs_btree_new_iroot, following the old naming but
making it clear we're dealing with the root in inode case here, and to
avoid confusion with xfs_btree_new_root which is used for the not inode
rooted case.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32201a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] implement semi-generic xfs_btree_new_root
Christoph Hellwig [Thu, 30 Oct 2008 05:57:16 +0000 (16:57 +1100)]
[XFS] implement semi-generic xfs_btree_new_root

From: Dave Chinner <dgc@sgi.com>

Add a xfs_btree_new_root helper for the alloc and ialloc btrees. The bmap
btree needs it's own version and is not converted.

[hch: split out from bigger patch and minor adaptions]

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32200a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] implement generic xfs_btree_split
Christoph Hellwig [Thu, 30 Oct 2008 05:57:03 +0000 (16:57 +1100)]
[XFS] implement generic xfs_btree_split

Make the btree split code generic. Based on a patch from David Chinner
with lots of changes to follow the original btree implementations more
closely. While this loses some of the generic helper routines for
inserting/moving/removing records it also solves some of the one off bugs
in the original code and makes it easier to verify.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32198a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] implement generic xfs_btree_lshift
Christoph Hellwig [Thu, 30 Oct 2008 05:56:53 +0000 (16:56 +1100)]
[XFS] implement generic xfs_btree_lshift

Make the btree left shift code generic. Based on a patch from David
Chinner with lots of changes to follow the original btree implementations
more closely. While this loses some of the generic helper routines for
inserting/moving/removing records it also solves some of the one off bugs
in the original code and makes it easier to verify.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32197a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] implement generic xfs_btree_rshift
Christoph Hellwig [Thu, 30 Oct 2008 05:56:43 +0000 (16:56 +1100)]
[XFS] implement generic xfs_btree_rshift

Make the btree right shift code generic. Based on a patch from David
Chinner with lots of changes to follow the original btree implementations
more closely. While this loses some of the generic helper routines for
inserting/moving/removing records it also solves some of the one off bugs
in the original code and makes it easier to verify.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32196a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] implement generic xfs_btree_update
Christoph Hellwig [Thu, 30 Oct 2008 05:56:32 +0000 (16:56 +1100)]
[XFS] implement generic xfs_btree_update

From: Dave Chinner <dgc@sgi.com>

The most complicated part here is the lastrec tracking for the alloc
btree. Most logic is in the update_lastrec method which has to do some
hopefully good enough dirty magic to maintain it.

[hch: split out from bigger patch and a rework of the lastrec

logic]

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32194a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] implement generic xfs_btree_updkey
Christoph Hellwig [Thu, 30 Oct 2008 05:56:22 +0000 (16:56 +1100)]
[XFS] implement generic xfs_btree_updkey

From: Dave Chinner <dgc@sgi.com>

Note that there are many > 80 char lines introduced due to the
xfs_btree_key casts. But the places where this happens is throw-away code
once the whole btree code gets merged into a common implementation.

The same is true for the temporary xfs_alloc_log_keys define to the new
name. All old users will be gone after a few patches.

[hch: split out from bigger patch and minor adaptions]

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32193a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] implement generic xfs_btree_lookup
Christoph Hellwig [Thu, 30 Oct 2008 05:56:09 +0000 (16:56 +1100)]
[XFS] implement generic xfs_btree_lookup

From: Dave Chinner <dgc@sgi.com>

[hch: split out from bigger patch and minor adaptions]

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32192a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] implement generic xfs_btree_decrement
Christoph Hellwig [Thu, 30 Oct 2008 05:55:58 +0000 (16:55 +1100)]
[XFS] implement generic xfs_btree_decrement

From: Dave Chinner <dgc@sgi.com>

[hch: split out from bigger patch and minor adaptions]

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32191a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] implement generic xfs_btree_increment
Christoph Hellwig [Thu, 30 Oct 2008 05:55:45 +0000 (16:55 +1100)]
[XFS] implement generic xfs_btree_increment

From: Dave Chinner <dgc@sgi.com>

Because this is the first major generic btree routine this patch includes
some infrastrucure, first a few routines to deal with a btree block that
can be either in short or long form, second xfs_btree_read_buf_block,
which is the new central routine to read a btree block given a cursor, and
third the new xfs_btree_ptr_addr routine to calculate the address for a
given btree pointer record.

[hch: split out from bigger patch and minor adaptions]

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32190a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] add helpers for addressing entities inside a btree block
Christoph Hellwig [Thu, 30 Oct 2008 05:55:34 +0000 (16:55 +1100)]
[XFS] add helpers for addressing entities inside a btree block

Add new helpers in xfs_btree.c to find the record, key and block pointer
entries inside a btree block. To implement this genericly the
->get_maxrecs methods and two new xfs_btree_ops entries for the key and
record sizes are used. Also add a big comment describing how the
addressing inside a btree block works.

Note that these helpers are unused until users are introduced in the next
patches and this patch will thus cause some harmless compiler warnings.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32189a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] add get_maxrecs btree operation
Christoph Hellwig [Thu, 30 Oct 2008 05:55:23 +0000 (16:55 +1100)]
[XFS] add get_maxrecs btree operation

Factor xfs_btree_maxrecs into a per-btree operation.

The get_maxrecs method is based on a patch from Dave Chinner.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32188a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] make btree tracing generic
Christoph Hellwig [Thu, 30 Oct 2008 05:55:13 +0000 (16:55 +1100)]
[XFS] make btree tracing generic

Make the existing bmap btree tracing generic so that it applies to all
btree types.

Some fragments lifted from a patch by Dave Chinner.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32187a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] add new btree statistics
David Chinner [Thu, 30 Oct 2008 05:55:03 +0000 (16:55 +1100)]
[XFS] add new btree statistics

From: Dave Chinner <dgc@sgi.com>

Introduce statistics coverage of all the btrees and cover all the btree
operations, not just some.

Invaluable for determining test code coverage of all the btree
operations....

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32184a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
16 years ago[XFS] refactor btree validation helpers
Christoph Hellwig [Thu, 30 Oct 2008 05:54:53 +0000 (16:54 +1100)]
[XFS] refactor btree validation helpers

Move the various btree validation helpers around in xfs_btree.c so that
they are close to each other and in common #ifdef DEBUG sections.

Also add a new xfs_btree_check_ptr helper to check a btree ptr that can be
either long or short form.

Split out from a bigger patch from Dave Chinner with various small changes
applied by me.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32183a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] refactor xfs_btree_readahead
Christoph Hellwig [Thu, 30 Oct 2008 05:54:43 +0000 (16:54 +1100)]
[XFS] refactor xfs_btree_readahead

From: Dave Chinner <dgc@sgi.com>

Refactor xfs_btree_readahead to make it more readable:

(a) remove the inline xfs_btree_readahead wrapper and move all checks out

of line into the main routine.

(b) factor out helpers for short/long form btrees

(c) move check for root in inodes from the callers into
xfs_btree_readahead

[hch: split out from a big patch and minor cleanups]

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32182a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] add a long pointers flag to xfs_btree_cur
Christoph Hellwig [Thu, 30 Oct 2008 05:54:33 +0000 (16:54 +1100)]
[XFS] add a long pointers flag to xfs_btree_cur

Add a flag to the xfs btree cursor when using long (64bit) block pointers
instead of checking btnum == XFS_BTNUM_BMAP.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32181a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] make btree root in inode support generic
Christoph Hellwig [Thu, 30 Oct 2008 05:54:22 +0000 (16:54 +1100)]
[XFS] make btree root in inode support generic

The bmap btree is rooted in the inode and not in a disk block. Make the
support for this feature more generic by adding a btree flag to for this
feature instead of relying on the XFS_BTNUM_BMAP btnum check.

Also clean up xfs_btree_get_block where this new flag is used.

Based upon a patch from Dave Chinner.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32180a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] add generic btree types
Christoph Hellwig [Thu, 30 Oct 2008 05:54:12 +0000 (16:54 +1100)]
[XFS] add generic btree types

Add generic union types for btree pointers, keys and records. The generic
btree pointer contains either a 32 and 64bit big endian scalar for short
and long form btrees, and the key and record contain the relevant type for
each possible btree.

Split out from a bigger patch from Dave Chinner and simplified a little
further.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32178a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] split up xfs_btree_init_cursor
Christoph Hellwig [Thu, 30 Oct 2008 05:53:59 +0000 (16:53 +1100)]
[XFS] split up xfs_btree_init_cursor

xfs_btree_init_cursor contains close to little shared code for the
different btrees and will get even more non-common code in the future.
Split it up into one routine per btree type.

Because xfs_btree_dup_cursor needs to call the init routine for a generic
btree cursor add a new btree operation vector that contains a dup_cursor
method that initializes a new cursor based on an existing one.

The btree operations vector is based on an idea and code from Dave Chinner
and will grow more entries later during this series.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32176a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] kill struct xfs_btree_hdr
Christoph Hellwig [Thu, 30 Oct 2008 05:53:47 +0000 (16:53 +1100)]
[XFS] kill struct xfs_btree_hdr

This type is only embedded in struct xfs_btree_block and never used
directly. By moving the fields directly into struct xfs_btree_block a lot
of the macros for struct xfs_btree_sblock and struct xfs_btree_lblock can
be used for struct xfs_btree_block too now which helps greatly with some
of the migrations during implementing the generic btree code.

SGI-PV: 985583

SGI-Modid: xfs-linux-melb:xfs-kern:32174a

Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Bill O'Donnell <billodo@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] Unlock inode before calling xfs_idestroy()
Lachlan McIlroy [Thu, 30 Oct 2008 05:53:38 +0000 (16:53 +1100)]
[XFS] Unlock inode before calling xfs_idestroy()

Lock debugging reported the ilock was being destroyed without being
unlocked. We don't need to lock the inode until we are going to insert it
into the radix tree.

SGI-PV: 987246

SGI-Modid: xfs-linux-melb:xfs-kern:32159a

Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years ago[XFS] Fix use-after-free with log and quotas
Lachlan McIlroy [Thu, 30 Oct 2008 05:53:25 +0000 (16:53 +1100)]
[XFS] Fix use-after-free with log and quotas

Destroying the quota stuff on unmount can access the log - ie
XFS_QM_DONE() ends up in xfs_dqunlock() which calls
xfs_trans_unlocked_item() and then xfs_log_move_tail(). By this time the
log has already been destroyed. Just move the cleanup of the quota code
earlier in xfs_unmountfs() before the call to xfs_log_unmount(). Moving
XFS_QM_DONE() up near XFS_QM_DQPURGEALL() seems like a good spot.

SGI-PV: 987086

SGI-Modid: xfs-linux-melb:xfs-kern:32148a

Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Peter Leckie <pleckie@sgi.com>
16 years ago[XFS] Remove final remnants of dirv1 macros and other stuff
Barry Naujok [Thu, 30 Oct 2008 05:52:35 +0000 (16:52 +1100)]
[XFS] Remove final remnants of dirv1 macros and other stuff

SGI-PV: 981498

SGI-Modid: xfs-linux-melb:xfs-kern:32002a

Signed-off-by: Barry Naujok <bnaujok@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Use xfs_idestroy() to cleanup an inode.
Lachlan McIlroy [Thu, 30 Oct 2008 05:50:35 +0000 (16:50 +1100)]
[XFS] Use xfs_idestroy() to cleanup an inode.

SGI-PV: 981498

SGI-Modid: xfs-linux-melb:xfs-kern:31927a

Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: David Chinner <david@fromorbit.com>
16 years ago[XFS] Remove kmem_zone_t argument from xfs_inode_init_once()
Lachlan McIlroy [Thu, 30 Oct 2008 05:42:34 +0000 (16:42 +1100)]
[XFS] Remove kmem_zone_t argument from xfs_inode_init_once()

kmem cache constructor no longer takes a kmem_zone_t argument.

SGI-PV: 957103

SGI-Modid: xfs-linux-melb:xfs-kern:32254a

Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
16 years ago[XFS] Make use of the init-once slab optimisation.
David Chinner [Thu, 30 Oct 2008 05:11:59 +0000 (16:11 +1100)]
[XFS] Make use of the init-once slab optimisation.

To avoid having to initialise some fields of the XFS inode on every
allocation, we can use the slab init-once feature to initialise them. All
we have to guarantee is that when we free the inode, all it's entries are
in the initial state. Add asserts where possible to ensure debug kernels
check this initial state before freeing and after allocation.

SGI-PV: 981498

SGI-Modid: xfs-linux-melb:xfs-kern:31925a

Signed-off-by: David Chinner <david@fromorbit.com>
Signed-off-by: Lachlan McIlroy <lachlan@sgi.com>
Signed-off-by: Christoph Hellwig <hch@infradead.org>
16 years agoMerge branch 'tracing-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Tue, 28 Oct 2008 16:52:25 +0000 (09:52 -0700)]
Merge branch 'tracing-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'tracing-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip: (31 commits)
  ftrace: fix current_tracer error return
  tracing: fix a build error on alpha
  ftrace: use a real variable for ftrace_nop in x86
  tracing/ftrace: make boot tracer select the sched_switch tracer
  tracepoint: check if the probe has been registered
  asm-generic: define DIE_OOPS in asm-generic
  trace: fix printk warning for u64
  ftrace: warning in kernel/trace/ftrace.c
  ftrace: fix build failure
  ftrace, powerpc, sparc64, x86: remove notrace from arch ftrace file
  ftrace: remove ftrace hash
  ftrace: remove mcount set
  ftrace: remove daemon
  ftrace: disable dynamic ftrace for all archs that use daemon
  ftrace: add ftrace warn on to disable ftrace
  ftrace: only have ftrace_kill atomic
  ftrace: use probe_kernel
  ftrace: comment arch ftrace code
  ftrace: return error on failed modified text.
  ftrace: dynamic ftrace process only text section
  ...

16 years agoMerge branch 'kvm-updates/2.6.28' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 28 Oct 2008 16:50:11 +0000 (09:50 -0700)]
Merge branch 'kvm-updates/2.6.28' of git://git./linux/kernel/git/avi/kvm

* 'kvm-updates/2.6.28' of git://git.kernel.org/pub/scm/linux/kernel/git/avi/kvm:
  KVM: ia64: Makefile fix for forcing to re-generate asm-offsets.h
  KVM: Future-proof device assignment ABI
  KVM: ia64: Fix halt emulation logic
  KVM: Fix guest shared interrupt with in-kernel irqchip
  KVM: MMU: sync root on paravirt TLB flush

16 years agoMerge branch 'core-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Tue, 28 Oct 2008 16:49:27 +0000 (09:49 -0700)]
Merge branch 'core-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'core-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  lockdep: fix irqs on/off ip tracing
  lockdep: minor fix for debug_show_all_locks()
  x86: restore the old swiotlb alloc_coherent behavior
  x86: use GFP_DMA for 24bit coherent_dma_mask
  swiotlb: remove panic for alloc_coherent failure
  xen: compilation fix of drivers/xen/events.c on IA64
  xen: portability clean up and some minor clean up for xencomm.c
  xen: don't reload cr3 on suspend
  kernel/resource: fix reserve_region_with_split() section mismatch
  printk: remove unused code from kernel/printk.c

16 years agoMerge branch 'irq-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 28 Oct 2008 16:48:25 +0000 (09:48 -0700)]
Merge branch 'irq-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'irq-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  irq: make variable static

16 years agoMerge branch 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Tue, 28 Oct 2008 16:46:20 +0000 (09:46 -0700)]
Merge branch 'sched-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'sched-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  sched: fix documentation reference for sched_min_granularity_ns
  sched: virtual time buddy preemption
  sched: re-instate vruntime based wakeup preemption
  sched: weaken sync hint
  sched: more accurate min_vruntime accounting
  sched: fix a find_busiest_group buglet
  sched: add CONFIG_SMP consistency

16 years agoMerge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 28 Oct 2008 16:45:31 +0000 (09:45 -0700)]
Merge branch 'x86-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  x86, memory hotplug: remove wrong -1 in calling init_memory_mapping()
  x86: keep the /proc/meminfo page count correct
  x86/uv: memory allocation at initialization
  xen: fix Xen domU boot with batched mprotect

16 years agoMerge branch 'for-linus' of git://git390.osdl.marist.edu/pub/scm/linux-2.6
Linus Torvalds [Tue, 28 Oct 2008 16:44:59 +0000 (09:44 -0700)]
Merge branch 'for-linus' of git://git390.osdl.marist.edu/linux-2.6

* 'for-linus' of git://git390.osdl.marist.edu/pub/scm/linux-2.6:
  [S390] s390: Fix build for !CONFIG_S390_GUEST + CONFIG_VIRTIO_CONSOLE
  [S390] No more 4kb stacks.
  [S390] Change default IPL method to IPL_VM.
  [S390] tape: disable interrupts in tape_open and tape_release
  [S390] appldata: unsigned ops->size cannot be negative
  [S390] tape block: complete request with correct locking
  [S390] Fix sysdev class file creation.
  [S390] pgtables: Fix race in enable_sie vs. page table ops
  [S390] qdio: remove incorrect memset
  [S390] qdio: prevent double qdio shutdown in case of I/O errors

16 years agoMerge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzi...
Linus Torvalds [Tue, 28 Oct 2008 16:42:48 +0000 (09:42 -0700)]
Merge branch 'upstream-linus' of git://git./linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
  libata: ahci enclosure management bit mask
  libata: ahci enclosure management led sync
  pata_ninja32: suspend/resume support
  libata: Fix LBA48 on pata_it821x RAID volumes.
  libata: clear saved xfer_mode and ncq_enabled on device detach
  sata_sil24: configure max read request size to 4k
  libata: add missing kernel-doc
  libata: fix device iteration bugs
  ahci: Add support for Promise PDC42819
  ata: Switch all my stuff to a common address

16 years agoftrace: fix current_tracer error return
Steven Rostedt [Tue, 28 Oct 2008 14:44:24 +0000 (10:44 -0400)]
ftrace: fix current_tracer error return

The commit (in linux-tip) c2931e05ec5965597cbfb79ad332d4a29aeceb23
 ( ftrace: return an error when setting a nonexistent tracer )
added useful code that would error when a bad tracer was written into
the current_tracer file.

But this had a bug if the amount written was more than the amount read by
that code. The first iteration would set the tracer correctly, but since
it did not consume the rest of what was written (usually whitespace), the
userspace utility would continue to write what was not consumed. This
second iteration would fail to find a tracer and return -EINVAL. Funny
thing is that the tracer would have already been set.

This patch just consumes all the data that is written to the file.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years agoKVM: ia64: Makefile fix for forcing to re-generate asm-offsets.h
Xiantao Zhang [Thu, 23 Oct 2008 02:37:06 +0000 (10:37 +0800)]
KVM: ia64: Makefile fix for forcing to re-generate asm-offsets.h

To avoid using stale asm-offsets.h.

Signed-off-by: Xiantao Zhang <xiantao.zhang@intel.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
16 years agoKVM: Future-proof device assignment ABI
Avi Kivity [Sun, 19 Oct 2008 14:39:45 +0000 (16:39 +0200)]
KVM: Future-proof device assignment ABI

Reserve some space so we can add more data.

Signed-off-by: Avi Kivity <avi@qumranet.com>
16 years agoKVM: ia64: Fix halt emulation logic
Xiantao Zhang [Thu, 16 Oct 2008 07:58:15 +0000 (15:58 +0800)]
KVM: ia64: Fix halt emulation logic

Common halt logic was changed by x86 and did not update ia64.  This patch
updates halt for ia64.

Fixes a regression causing guests to hang with more than 2 vcpus.

Signed-off-by: Xiantao Zhang <xiantao.zhang@intel.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
16 years agoKVM: Fix guest shared interrupt with in-kernel irqchip
Sheng Yang [Wed, 15 Oct 2008 12:15:06 +0000 (20:15 +0800)]
KVM: Fix guest shared interrupt with in-kernel irqchip

Every call of kvm_set_irq() should offer an irq_source_id, which is
allocated by kvm_request_irq_source_id(). Based on irq_source_id, we
identify the irq source and implement logical OR for shared level
interrupts.

The allocated irq_source_id can be freed by kvm_free_irq_source_id().

Currently, we support at most sizeof(unsigned long) different irq sources.

[Amit: - rebase to kvm.git HEAD
       - move definition of KVM_USERSPACE_IRQ_SOURCE_ID to common file
       - move kvm_request_irq_source_id to the update_irq ioctl]

[Xiantao: - Add kvm/ia64 stuff and make it work for kvm/ia64 guests]

Signed-off-by: Sheng Yang <sheng@linux.intel.com>
Signed-off-by: Amit Shah <amit.shah@redhat.com>
Signed-off-by: Xiantao Zhang <xiantao.zhang@intel.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
16 years agoKVM: MMU: sync root on paravirt TLB flush
Marcelo Tosatti [Wed, 15 Oct 2008 09:45:08 +0000 (07:45 -0200)]
KVM: MMU: sync root on paravirt TLB flush

The pvmmu TLB flush handler should request a root sync, similarly to
a native read-write CR3.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
16 years agolockdep: fix irqs on/off ip tracing
Heiko Carstens [Tue, 28 Oct 2008 10:14:58 +0000 (11:14 +0100)]
lockdep: fix irqs on/off ip tracing

Impact: fix lockdep lock-api-caller output when irqsoff tracing is enabled

81d68a96 "ftrace: trace irq disabled critical timings" added wrappers around
trace_hardirqs_on/off_caller. However these functions use
__builtin_return_address(0) to figure out which function actually disabled
or enabled irqs. The result is that we save the ips of trace_hardirqs_on/off
instead of the real caller. Not very helpful.

However since the patch from Steven the ip already gets passed. So use that
and get rid of __builtin_return_address(0) in these two functions.

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
16 years ago[S390] s390: Fix build for !CONFIG_S390_GUEST + CONFIG_VIRTIO_CONSOLE
Christian Borntraeger [Tue, 28 Oct 2008 10:10:22 +0000 (11:10 +0100)]
[S390] s390: Fix build for !CONFIG_S390_GUEST + CONFIG_VIRTIO_CONSOLE

The s390 kernel does not compile if virtio console is enabled, but guest
support is disabled:

  LD      .tmp_vmlinux1
arch/s390/kernel/built-in.o: In function `setup_arch':
/space/linux-2.5/arch/s390/kernel/setup.c:773: undefined reference to
`s390_virtio_console_init'

The fix is related to
commit 99e65c92f2bbf84f43766a8bf701e36817d62822
Author: Christian Borntraeger <borntraeger@de.ibm.com>
Date:   Fri Jul 25 15:50:04 2008 +0200
    KVM: s390: Fix guest kconfig

Which changed the build process to build kvm_virtio.c only if CONFIG_S390_GUEST
is set. We must ifdef the prototype in the header file accordingly.

Reported-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] No more 4kb stacks.
Heiko Carstens [Tue, 28 Oct 2008 10:10:21 +0000 (11:10 +0100)]
[S390] No more 4kb stacks.

We got a stack overflow with a small stack configuration on a 32 bit
system. It just looks like as 4kb isn't enough and too dangerous.
So lets get rid of 4kb stacks on 32 bit.

But one thing I completely dislike about the call trace below is that
just for debugging or tracing purposes sprintf gets called (cio_start_key):

/* process condition code */
sprintf(dbf_txt, "ccode:%d", ccode);
CIO_TRACE_EVENT(4, dbf_txt);

But maybe its just me who thinks that this could be done better.

    <4>Kernel stack overflow.
    <4>Modules linked in: dm_multipath sunrpc bonding qeth_l2 dm_mod qeth ccwgroup vmur
    <4>CPU: 1 Not tainted 2.6.27-30.x.20081015-s390default #1
    <4>Process httpd (pid: 3807, task: 20ae2df8, ksp: 1666fb78)
    <4>Krnl PSW : 040c0000 8027098a (number+0xe/0x348)
    <4>           R:0 T:1 IO:0 EX:0 Key:0 M:1 W:0 P:0 AS:0 CC:0 PM:0
    <4>Krnl GPRS: 00d43318 0027097c 1666f277 9666f270
    <4>           00000000 00000000 0000000a ffffffff
    <4>           9666f270 1666f228 1666f277 1666f098
    <4>           00000002 80270982 80271016 1666f098
    <4>Krnl Code: 8027097ef0340dd0a7f1 srp 3536(4,%r0),2033(%r10),4
    <4>           80270984: 0f00 clcl %r0,%r0
    <4>           80270986a7840001 brc 8,80270988
    <4>          >8027098a: 18ef lr %r14,%r15
    <4>           8027098ca7faff68 ahi %r15,-152
    <4>           80270990: 18bf lr %r11,%r15
    <4>           80270992: 18a2 lr %r10,%r2
    <4>           80270994: 1893 lr %r9,%r3

Modified calltrace with annotated stackframe size of each function:

stackframe size
    |
 0 304 vsnprintf+850 [0x271016]
 1  72 sprintf+74 [0x271522]
 2  56 cio_start_key+262 [0x2d4c16]
 3  56 ccw_device_start_key+222 [0x2dfe92]
 4  56 ccw_device_start+40 [0x2dff28]
 5  48 raw3215_start_io+104 [0x30b0f8]
 6  56 raw3215_write+494 [0x30ba0a]
 7  40 con3215_write+68 [0x30bafc]
 8  40 __call_console_drivers+146 [0x12b0fa]
 9  32 _call_console_drivers+102 [0x12b192]
10  64 release_console_sem+268 [0x12b614]
11 168 vprintk+462 [0x12bca6]
12  72 printk+68 [0x12bfd0]
13 256 __print_symbol+50 [0x15a882]
14  56 __show_trace+162 [0x103d06]
15  32 show_trace+224 [0x103e70]
16  48 show_stack+152 [0x103f20]
17  56 dump_stack+126 [0x104612]
18  96 __alloc_pages_internal+592 [0x175004]
19  80 cache_alloc_refill+776 [0x196f3c]
20  40 __kmalloc+258 [0x1972ae]
21  40 __alloc_skb+94 [0x328086]
22  32 pskb_copy+50 [0x328252]
23  32 skb_realloc_headroom+110 [0x328a72]
24 104 qeth_l2_hard_start_xmit+378 [0x7803bfde]
25  56 dev_hard_start_xmit+450 [0x32ef6e]
26  56 __qdisc_run+390 [0x3425d6]
27  48 dev_queue_xmit+410 [0x331e06]
28  40 ip_finish_output+308 [0x354ac8]
29  56 ip_output+218 [0x355b6e]
30  24 ip_local_out+56 [0x354584]
31 120 ip_queue_xmit+300 [0x355cec]
32  96 tcp_transmit_skb+812 [0x367da8]
33  40 tcp_push_one+158 [0x369fda]
34 112 tcp_sendmsg+852 [0x35d5a0]
35 240 sock_sendmsg+164 [0x32035c]
36  56 kernel_sendmsg+86 [0x32064a]
37  88 sock_no_sendpage+98 [0x322b22]
38 104 tcp_sendpage+70 [0x35cc1e]
39  48 sock_sendpage+74 [0x31eb66]
40  64 pipe_to_sendpage+102 [0x1c4b2e]
41  64 __splice_from_pipe+120 [0x1c5340]
42  72 splice_from_pipe+90 [0x1c57e6]
43  56 generic_splice_sendpage+38 [0x1c5832]
44  48 do_splice_from+104 [0x1c4c38]
45  48 direct_splice_actor+52 [0x1c4c88]
46  80 splice_direct_to_actor+180 [0x1c4f80]
47  72 do_splice_direct+70 [0x1c5112]
48  64 do_sendfile+360 [0x19de18]
49  72 sys_sendfile64+126 [0x19df32]
50 336 sysc_do_restart+18 [0x111a1a]

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] Change default IPL method to IPL_VM.
Heiko Carstens [Tue, 28 Oct 2008 10:10:20 +0000 (11:10 +0100)]
[S390] Change default IPL method to IPL_VM.

allyesconfig and allmodconfig built kernels have a tape IPL record.
A the vmreader record makes much more sense, since hardly anybody will
ever IPL a kernel from tape. So change the default.
As I side effect I can test these kernels without fiddling around with
the kernel config ;)

Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] tape: disable interrupts in tape_open and tape_release
Frank Munzert [Tue, 28 Oct 2008 10:10:19 +0000 (11:10 +0100)]
[S390] tape: disable interrupts in tape_open and tape_release

Get tape device lock with interrupts disabled. Otherwise lockdep will issue a
warning similar to:

=================================
[ INFO: inconsistent lock state ]
2.6.27 #1
---------------------------------
inconsistent {in-hardirq-W} -> {hardirq-on-W} usage.
vol_id/2903 [HC0[0]:SC0[0]:HE1:SE1] takes:
 (sch->lock){++..}, at: [<000003e00004c7a2>] tape_open+0x42/0x1a4 [tape]
{in-hardirq-W} state was registered at:
  [<000000000007ce5c>] __lock_acquire+0x894/0xa74
  [<000000000007d0ce>] lock_acquire+0x92/0xb8
  [<0000000000345154>] _spin_lock+0x5c/0x9c
  [<0000000000202264>] do_IRQ+0x124/0x1f0
  [<0000000000026610>] io_return+0x0/0x8

irq event stamp: 847
hardirqs last  enabled at (847): [<000000000007aca6>] trace_hardirqs_on+0x2a/0x38
hardirqs last disabled at (846): [<0000000000076ca2>] trace_hardirqs_off+0x2a/0x38
softirqs last  enabled at (0): [<000000000004909e>] copy_process+0x43e/0x11f4
softirqs last disabled at (0): [<0000000000000000>] 0x0

other info that might help us debug this:
1 lock held by vol_id/2903:
 #0:  (&bdev->bd_mutex){--..}, at: [<000000000010e0f4>] do_open+0x78/0x358

stack backtrace:
CPU: 1 Not tainted 2.6.27 #1},
Process vol_id (pid: 2903, task: 000000003d4c0000, ksp: 000000003d4e3b10)
0400000000000000 000000003d4e3830 0000000000000002 0000000000000000
       000000003d4e38d0 000000003d4e3848 000000003d4e3848 00000000000168a8
       0000000000000000 000000003d4e3b10 0000000000000000 0000000000000000
       000000003d4e3830 000000000000000c 000000003d4e3830 000000003d4e38a0
       000000000034aa98 00000000000168a8 000000003d4e3830 000000003d4e3880
Call Trace:
([<000000000001681c>] show_trace+0x138/0x158)
 [<0000000000016902>] show_stack+0xc6/0xf8
 [<00000000000170d4>] dump_stack+0xb0/0xc0
 [<0000000000078810>] print_usage_bug+0x1e8/0x228
 [<000000000007a71c>] mark_lock+0xb14/0xd24
 [<000000000007cd5a>] __lock_acquire+0x792/0xa74
 [<000000000007d0ce>] lock_acquire+0x92/0xb8
 [<0000000000345154>] _spin_lock+0x5c/0x9c
 [<000003e00004c7a2>] tape_open+0x42/0x1a4 [tape]
 [<000003e00005185c>] tapeblock_open+0x98/0xd0 [tape]

Signed-off-by: Frank Munzert <munzert@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] appldata: unsigned ops->size cannot be negative
Roel Kluin [Tue, 28 Oct 2008 10:10:18 +0000 (11:10 +0100)]
[S390] appldata: unsigned ops->size cannot be negative

unsigned ops->size cannot be negative

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] tape block: complete request with correct locking
Frank Munzert [Tue, 28 Oct 2008 10:10:17 +0000 (11:10 +0100)]
[S390] tape block: complete request with correct locking

__blk_end_request must be called with request queue lock held. We need to use
blk_end_request rather than  __blk_end_request.

Signed-off-by: Frank Munzert <munzert@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] Fix sysdev class file creation.
Heiko Carstens [Tue, 28 Oct 2008 10:10:16 +0000 (11:10 +0100)]
[S390] Fix sysdev class file creation.

Use sysdev_class_create_file() to create create sysdev class attributes
instead of sysfs_create_file(). Using sysfs_create_file() wasn't a very
good idea since the show and store functions have a different amount of
parameters for sysfs files and sysdev class files.
In particular the pointer to the buffer is the last argument and
therefore accesses to random memory regions happened.
Still worked surprisingly well until we got a kernel panic.

Cc: stable@kernel.org
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] pgtables: Fix race in enable_sie vs. page table ops
Christian Borntraeger [Tue, 28 Oct 2008 10:10:15 +0000 (11:10 +0100)]
[S390] pgtables: Fix race in enable_sie vs. page table ops

The current enable_sie code sets the mm->context.pgstes bit to tell
dup_mm that the new mm should have extended page tables. This bit is also
used by the s390 specific page table primitives to decide about the page
table layout - which means context.pgstes has two meanings. This can cause
any kind of bugs. For example  - e.g. shrink_zone can call
ptep_clear_flush_young while enable_sie is running. ptep_clear_flush_young
will test for context.pgstes. Since enable_sie changed that value of the old
struct mm without changing the page table layout ptep_clear_flush_young will
do the wrong thing.
The solution is to split pgstes into two bits
- one for the allocation
- one for the current state

Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
16 years ago[S390] qdio: remove incorrect memset
Jan Glauber [Tue, 28 Oct 2008 10:10:14 +0000 (11:10 +0100)]
[S390] qdio: remove incorrect memset

Remove the memset since zeroing the string is not needed and use
snprintf instead of sprintf.

Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>