Just minor housekeeping, a lot more functions can be trivially made
static; others could if we reordered things a bit...
Signed-off-by: NEric Sandeen <sandeen@sandeen.net>
Signed-off-by: NAlex Elder <aelder@sgi.com>

The swap extent ioctl passes in a target inode and a temporary inode
which are clearly named in the ioctl structure. The code then
assigns temp to target and vice versa, making it extremely difficult
to work out which inode is which later in the code.  Make this
consistent throughout the code.

Also make xfs_swap_extent static as there are no external users of
the function.
Signed-off-by: NDave Chinner <david@fromorbit.com>
Signed-off-by: NAlex Elder <aelder@sgi.com>

To be able to diagnose whether the swap extents function is
detecting compatible inode data fork configurations for swapping
extents, add tracing points to the code to allow us to see the
format of the inode forks before and after the swap.
Signed-off-by: NDave Chinner <david@fromorbit.com>
Signed-off-by: NAlex Elder <aelder@sgi.com>

When swapping extents, we can corrupt inodes by swapping data forks
that are in incompatible formats.  This is caused by the two indoes
having different fork offsets due to the presence of an attribute
fork on an attr2 filesystem.  xfs_fsr tries to be smart about
setting the fork offset, but the trick it plays only works on attr1
(old fixed format attribute fork) filesystems.

Changing the way xfs_fsr sets up the attribute fork will prevent
this situation from ever occurring, so in the kernel code we can get
by with a preventative fix - check that the data fork in the
defragmented inode is in a format valid for the inode it is being
swapped into.  This will lead to files that will silently and
potentially repeatedly fail defragmentation, so issue a warning to
the log when this particular failure occurs to let us know that
xfs_fsr needs updating/fixing.

To help identify how to improve xfs_fsr to avoid this issue, add
trace points for the inodes being swapped so that we can determine
why the swap was rejected and to confirm that the code is making the
right decisions and modifications when swapping forks.

A further complication is even when the swap is allowed to proceed
when the fork offset is different between the two inodes then value
for the maximum number of extents the data fork can hold can be
wrong. Make sure these are also set correctly after the swap occurs.
Signed-off-by: NDave Chinner <david@fromorbit.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAlex Elder <aelder@sgi.com>

When xfs_rtfind_forw() returns an error, the block is returned
uninitialised.  xfs_rtfree_range() is not checking the error return,
so could be using an uninitialised block number for modifying bitmap
summary info.

The problem was found by gcc when compiling the *userspace* libxfs
code - it is an copy of the kernel code with the exact same bug.
gcc gives an uninitialised variable warning on the userspace code
but not on the kernel code. You gotta love the consistency (Mmmm,
slightly chewy today!).
Signed-off-by: NDave Chinner <david@fromorbit.com>
Signed-off-by: NAlex Elder <aelder@sgi.com>

When reclaiming stale inodes, we need to guarantee that inodes are
unpinned before returning with a "clean" status. If we don't we can
reclaim inodes that are pinned, leading to use after free in the
transaction subsystem as transactions complete.
Signed-off-by: NDave Chinner <david@fromorbit.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAlex Elder <aelder@sgi.com>

lockdep complains about a the lock not being initialised as we do an
ASSERT based check that the lock is not held before we initialise it
to catch inodes freed with the lock held.

lockdep does this check for us in the lock initialisation code, so
remove the ASSERT to stop the lockdep warning.
Signed-off-by: NDave Chinner <david@fromorbit.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAlex Elder <aelder@sgi.com>

We cannot do direct inode reclaim without taking the flush lock to
ensure that we do not reclaim an inode under IO. We check the inode
is clean before doing direct reclaim, but this is not good enough
because the inode flush code marks the inode clean once it has
copied the in-core dirty state to the backing buffer.

It is the flush lock that determines whether the inode is still
under IO, even though it is marked clean, and the inode is still
required at IO completion so we can't reclaim it even though it is
clean in core. Hence the requirement that we need to take the flush
lock even on clean inodes because this guarantees that the inode
writeback IO has completed and it is safe to reclaim the inode.

With delayed write inode flushing, we coul dend up waiting a long
time on the flush lock even for a clean inode. The background
reclaim already handles this efficiently, so avoid all the problems
by killing the direct reclaim path altogether.
Signed-off-by: NDave Chinner <david@fromorbit.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAlex Elder <aelder@sgi.com>

The reclaim code will handle flushing of dirty inodes before reclaim
occurs, so avoid them when determining whether an inode is a
candidate for flushing to disk when walking the radix trees.  This
is based on a test patch from Christoph Hellwig.
Signed-off-by: NDave Chinner <david@fromorbit.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAlex Elder <aelder@sgi.com>

Make the inode tree reclaim walk exclusive to avoid races with
concurrent sync walkers and lookups. This is a version of a patch
posted by Christoph Hellwig that avoids all the code duplication.
Signed-off-by: NDave Chinner <david@fromorbit.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAlex Elder <aelder@sgi.com>

When we search for and find a busy extent during allocation we
force the log out to ensure the extent free transaction is on
disk before the allocation transaction. The current implementation
has a subtle bug in it--it does not handle multiple overlapping
ranges.

That is, if we free lots of little extents into a single
contiguous extent, then allocate the contiguous extent, the busy
search code stops searching at the first extent it finds that
overlaps the allocated range. It then uses the commit LSN of the
transaction to force the log out to.

Unfortunately, the other busy ranges might have more recent
commit LSNs than the first busy extent that is found, and this
results in xfs_alloc_search_busy() returning before all the
extent free transactions are on disk for the range being
allocated. This can lead to potential metadata corruption or
stale data exposure after a crash because log replay won't replay
all the extent free transactions that cover the allocation range.
Modified-by: NAlex Elder <aelder@sgi.com>

(Dropped the "found" argument from the xfs_alloc_busysearch trace
event.)
Signed-off-by: NDave Chinner <david@fromorbit.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAlex Elder <aelder@sgi.com>

Because inodes remain in cache much longer than inode buffers do
under memory pressure, we can get the situation where we have
stale, dirty inodes being reclaimed but the backing storage has
been freed.  Hence we should never, ever flush XFS_ISTALE inodes
to disk as there is no guarantee that the backing buffer is in
cache and still marked stale when the flush occurs.
Signed-off-by: NDave Chinner <david@fromorbit.com>
Signed-off-by: NAlex Elder <aelder@sgi.com>

We currently have some rather odd code in xfs_setattr for
updating the a/c/mtime timestamps:

 - first we do a non-transaction update if all three are updated
   together
 - second we implicitly update the ctime for various changes
   instead of relying on the ATTR_CTIME flag
 - third we set the timestamps to the current time instead of the
   arguments in the iattr structure in many cases.

This patch makes sure we update it in a consistent way:

 - always transactional
 - ctime is only updated if ATTR_CTIME is set or we do a size
   update, which is a special case
 - always to the times passed in from the caller instead of the
   current time

The only non-size caller of xfs_setattr that doesn't come from
the VFS is updated to set ATTR_CTIME and pass in a valid ctime
value.
Reported-by: NEric Blake <ebb9@byu.net>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAlex Elder <aelder@sgi.com>

Using DECLARE_EVENT_CLASS allows us to to use trace event code
instead of duplicating it in the binary.  This was not available
before 2.6.33 so it had to be done as a separate step once the
prerequisite was merged.

This only requires changes to xfs_trace.h and the results are
rather impressive:

hch@brick:~/work/linux-2.6/obj-kvm$ size fs/xfs/xfs.o*
text	   data	    bss	    dec	    hex	filename
 607732	  41884	   3616	 653232	  9f7b0	fs/xfs/xfs.o
1026732	  41884	   3808	1072424	 105d28	fs/xfs/xfs.o.old
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAlex Elder <aelder@sgi.com>

Randy Dunlap Reported printk() format-related warnings reported
on i386 builds in his environment.  Dave Chinner provided this
patch to eliminate them.

Signed-off by: Dave Chinner <david@fromorbit.com>
Acked-by: NRandy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: NAlex Elder <aelder@sgi.com>

After I_SYNC was split from I_LOCK the leftover is always used together with
I_NEW and thus superflous.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

The code in xfs_free_buf() only attempts to free the b_pages array if the
buffer is a page cache backed or page allocated buffer. The extra log buffer
that is used when the log wraps uses pages that are allocated to a different
log buffer, but it still has a b_pages array allocated when those pages
are associated to with the extra buffer in xfs_buf_associate_memory.

Hence we need to always attempt to free the b_pages array when tearing
down a buffer, not just on buffers that are explicitly marked as page bearing
buffers. This fixes a leak detected by the kernel memory leak code.
Signed-off-by: NDave Chinner <david@fromorbit.com>
Signed-off-by: NAlex Elder <aelder@sgi.com>

For a long time we've always stored bmap btree records in the 64bit format,
so kill off the dead 32bit type, and make sure the 64bit type is named just
xfs_bmbt_rec everywhere, without any size postfix.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NEric Sandeen <sandeen@redhat.com>
Signed-off-by: NAlex Elder <aelder@sgi.com>

Change all async metadata buffers to use [READ|WRITE]_META I/O types
so that the I/O doesn't get issued immediately. This allows merging of
adjacent metadata requests but still prioritises them over bulk data.
This shows a 10-15% improvement in sequential create speed of small
files.

Don't include the log buffers in this classification - leave them as
sync types so they are issued immediately.
Signed-off-by: NDave Chinner <dgc@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAlex Elder <aelder@sgi.com>

Add an assert for inodes not added to the inode cache in xfs_ireclaim,
to make sure we're not going to introduce something like the
famous nfsd inode cache bug again.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAlex Elder <aelder@sgi.com>

Currently the locking in blockdev_direct_IO is a mess, we have three different
locking types and very confusing checks for some of them.  The most
complicated one is DIO_OWN_LOCKING for reads, which happens to not actually be
used.

This patch gets rid of the DIO_OWN_LOCKING - as mentioned above the read case
is unused anyway, and the write side is almost identical to DIO_NO_LOCKING.
The difference is that DIO_NO_LOCKING always sets the create argument for
the get_blocks callback to zero, but we can easily move that to the actual
get_blocks callbacks.  There are four users of the DIO_NO_LOCKING mode:
gfs already ignores the create argument and thus is fine with the new
version, ocfs2 only errors out if create were ever set, and we can remove
this dead code now, the block device code only ever uses create for an
error message if we are fully beyond the device which can never happen,
and last but not least XFS will need the new behavour for writes.

Now we can replace the lock_type variable with a flags one, where no flag
means the DIO_NO_LOCKING behaviour and DIO_LOCKING is kept as the first
flag.  Separate out the check for not allowing to fill holes into a separate
flag, although for now both flags always get set at the same time.

Also revamp the documentation of the locking scheme to actually make sense.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Add a flags argument to struct xattr_handler and pass it to all xattr
handler methods.  This allows using the same methods for multiple
handlers, e.g. for the ACL methods which perform exactly the same action
for the access and default ACLs, just using a different underlying
attribute.  With a little more groundwork it'll also allow sharing the
methods for the regular user/trusted/secure handlers in extN, ocfs2 and
jffs2 like it's already done for xfs in this patch.

Also change the inode argument to the handlers to a dentry to allow
using the handlers mechnism for filesystems that require it later,
e.g. cifs.

[with GFS2 bits updated by Steven Whitehouse <swhiteho@redhat.com>]
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NJames Morris <jmorris@namei.org>
Acked-by: NJoel Becker <joel.becker@oracle.com>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Currently the locking in blockdev_direct_IO is a mess, we have three
different locking types and very confusing checks for some of them.  The
most complicated one is DIO_OWN_LOCKING for reads, which happens to not
actually be used.

This patch gets rid of the DIO_OWN_LOCKING - as mentioned above the read
case is unused anyway, and the write side is almost identical to
DIO_NO_LOCKING.  The difference is that DIO_NO_LOCKING always sets the
create argument for the get_blocks callback to zero, but we can easily
move that to the actual get_blocks callbacks.  There are four users of the
DIO_NO_LOCKING mode: gfs already ignores the create argument and thus is
fine with the new version, ocfs2 only errors out if create were ever set,
and we can remove this dead code now, the block device code only ever uses
create for an error message if we are fully beyond the device which can
never happen, and last but not least XFS will need the new behavour for
writes.

Now we can replace the lock_type variable with a flags one, where no flag
means the DIO_NO_LOCKING behaviour and DIO_LOCKING is kept as the first
flag.  Separate out the check for not allowing to fill holes into a
separate flag, although for now both flags always get set at the same
time.

Also revamp the documentation of the locking scheme to actually make
sense.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
Cc: Jens Axboe <jens.axboe@oracle.com>
Cc: Zach Brown <zach.brown@oracle.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Alex Elder <aelder@sgi.com>
Cc: Mark Fasheh <mfasheh@suse.com>
Cc: Joel Becker <joel.becker@oracle.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: NJoe Perches <joe@perches.com>
Acked-by: NAlex Elder <aelder@sgi.com>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Convert the old xfs tracing support that could only be used with the
out of tree kdb and xfsidbg patches to use the generic event tracer.

To use it make sure CONFIG_EVENT_TRACING is enabled and then enable
all xfs trace channels by:

   echo 1 > /sys/kernel/debug/tracing/events/xfs/enable

or alternatively enable single events by just doing the same in one
event subdirectory, e.g.

   echo 1 > /sys/kernel/debug/tracing/events/xfs/xfs_ihold/enable

or set more complex filters, etc. In Documentation/trace/events.txt
all this is desctribed in more detail.  To reads the events do a

   cat /sys/kernel/debug/tracing/trace

Compared to the last posting this patch converts the tracing mostly to
the one tracepoint per callsite model that other users of the new
tracing facility also employ.  This allows a very fine-grained control
of the tracing, a cleaner output of the traces and also enables the
perf tool to use each tracepoint as a virtual performance counter,
     allowing us to e.g. count how often certain workloads git various
     spots in XFS.  Take a look at

    http://lwn.net/Articles/346470/

for some examples.

Also the btree tracing isn't included at all yet, as it will require
additional core tracing features not in mainline yet, I plan to
deliver it later.

And the really nice thing about this patch is that it actually removes
many lines of code while adding this nice functionality:

 fs/xfs/Makefile                |    8
 fs/xfs/linux-2.6/xfs_acl.c     |    1
 fs/xfs/linux-2.6/xfs_aops.c    |   52 -
 fs/xfs/linux-2.6/xfs_aops.h    |    2
 fs/xfs/linux-2.6/xfs_buf.c     |  117 +--
 fs/xfs/linux-2.6/xfs_buf.h     |   33
 fs/xfs/linux-2.6/xfs_fs_subr.c |    3
 fs/xfs/linux-2.6/xfs_ioctl.c   |    1
 fs/xfs/linux-2.6/xfs_ioctl32.c |    1
 fs/xfs/linux-2.6/xfs_iops.c    |    1
 fs/xfs/linux-2.6/xfs_linux.h   |    1
 fs/xfs/linux-2.6/xfs_lrw.c     |   87 --
 fs/xfs/linux-2.6/xfs_lrw.h     |   45 -
 fs/xfs/linux-2.6/xfs_super.c   |  104 ---
 fs/xfs/linux-2.6/xfs_super.h   |    7
 fs/xfs/linux-2.6/xfs_sync.c    |    1
 fs/xfs/linux-2.6/xfs_trace.c   |   75 ++
 fs/xfs/linux-2.6/xfs_trace.h   | 1369 +++++++++++++++++++++++++++++++++++++++++
 fs/xfs/linux-2.6/xfs_vnode.h   |    4
 fs/xfs/quota/xfs_dquot.c       |  110 ---
 fs/xfs/quota/xfs_dquot.h       |   21
 fs/xfs/quota/xfs_qm.c          |   40 -
 fs/xfs/quota/xfs_qm_syscalls.c |    4
 fs/xfs/support/ktrace.c        |  323 ---------
 fs/xfs/support/ktrace.h        |   85 --
 fs/xfs/xfs.h                   |   16
 fs/xfs/xfs_ag.h                |   14
 fs/xfs/xfs_alloc.c             |  230 +-----
 fs/xfs/xfs_alloc.h             |   27
 fs/xfs/xfs_alloc_btree.c       |    1
 fs/xfs/xfs_attr.c              |  107 ---
 fs/xfs/xfs_attr.h              |   10
 fs/xfs/xfs_attr_leaf.c         |   14
 fs/xfs/xfs_attr_sf.h           |   40 -
 fs/xfs/xfs_bmap.c              |  507 +++------------
 fs/xfs/xfs_bmap.h              |   49 -
 fs/xfs/xfs_bmap_btree.c        |    6
 fs/xfs/xfs_btree.c             |    5
 fs/xfs/xfs_btree_trace.h       |   17
 fs/xfs/xfs_buf_item.c          |   87 --
 fs/xfs/xfs_buf_item.h          |   20
 fs/xfs/xfs_da_btree.c          |    3
 fs/xfs/xfs_da_btree.h          |    7
 fs/xfs/xfs_dfrag.c             |    2
 fs/xfs/xfs_dir2.c              |    8
 fs/xfs/xfs_dir2_block.c        |   20
 fs/xfs/xfs_dir2_leaf.c         |   21
 fs/xfs/xfs_dir2_node.c         |   27
 fs/xfs/xfs_dir2_sf.c           |   26
 fs/xfs/xfs_dir2_trace.c        |  216 ------
 fs/xfs/xfs_dir2_trace.h        |   72 --
 fs/xfs/xfs_filestream.c        |    8
 fs/xfs/xfs_fsops.c             |    2
 fs/xfs/xfs_iget.c              |  111 ---
 fs/xfs/xfs_inode.c             |   67 --
 fs/xfs/xfs_inode.h             |   76 --
 fs/xfs/xfs_inode_item.c        |    5
 fs/xfs/xfs_iomap.c             |   85 --
 fs/xfs/xfs_iomap.h             |    8
 fs/xfs/xfs_log.c               |  181 +----
 fs/xfs/xfs_log_priv.h          |   20
 fs/xfs/xfs_log_recover.c       |    1
 fs/xfs/xfs_mount.c             |    2
 fs/xfs/xfs_quota.h             |    8
 fs/xfs/xfs_rename.c            |    1
 fs/xfs/xfs_rtalloc.c           |    1
 fs/xfs/xfs_rw.c                |    3
 fs/xfs/xfs_trans.h             |   47 +
 fs/xfs/xfs_trans_buf.c         |   62 -
 fs/xfs/xfs_vnodeops.c          |    8
 70 files changed, 2151 insertions(+), 2592 deletions(-)
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAlex Elder <aelder@sgi.com>

Change the xfs_iext_insert / xfs_iext_remove prototypes to pass more
information which will allow pushing the trace points from the callers
into those functions.  This includes folding the whichfork information
into the state variable to minimize the addition stack footprint.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAlex Elder <aelder@sgi.com>

Cleanup the extent state macros in the bmap code to use one common set of
flags that we can pass to the tracing code later and remove a lot of the
macro obsfucation.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAlex Elder <aelder@sgi.com>

Noticed that through glibc fallocate would return 28 rather than -1
and errno = 28 for ENOSPC. The xfs routines uses XFS_ERROR format
positive return error codes while the syscalls use negative return
codes.  Fixup the two cases in xfs_vn_fallocate syscall to convert to
negative.
Signed-off-by: NJason Gunthorpe <jgunthorpe@obsidianresearch.com>
Reviewed-by: NEric Sandeen <sandeen@sandeen.net>
Signed-off-by: NAlex Elder <aelder@sgi.com>

Stop the flag saving as we never mangle those in the unmount path, and
hide all the weird arguents to the dmapi code inside the
XFS_SEND_PREUNMOUNT / XFS_SEND_UNMOUNT macros.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <david@fromorbit.com>
Signed-off-by: NAlex Elder <aelder@sgi.com>

xfs_iget_cache_miss does not get called with the pag_ici_lock held, so
the __releases annotation is incorrect.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <david@fromorbit.com>
Signed-off-by: NAlex Elder <aelder@sgi.com>

Remove our own STATIC_INLINE macro.  For small function inside
implementation files just use STATIC and let gcc inline it, and for
those in headers do the normal static inline - they are all small
enough to be inlined for debug builds, too.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <david@fromorbit.com>
Signed-off-by: NAlex Elder <aelder@sgi.com>

This function is too large to efficiently be inlined.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <david@fromorbit.com>
Signed-off-by: NAlex Elder <aelder@sgi.com>

Using a totally different name for the low-level get operation does
not fit the _int convention used in the rest of the attr code, so
rename it.

While we're at it also fix the prototype to use the normal convention
and mark it static as it's never used outside of xfs_attr.c.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <david@fromorbit.com>
Reviewed-by: NEric Sandeen <sandeen@sandeen.net>
Signed-off-by: NAlex Elder <aelder@sgi.com>

Currently the low-level buffer cache interfaces are highly confusing
as we have a _flags variant of each that does actually respect the
flags, and one without _flags which has a flags argument that gets
ignored and overriden with a default set.  Given that very few places
use the default arguments get rid of the duplication and convert all
callers to pass the flags explicitly.  Also remove the now confusing
_flags postfix.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <david@fromorbit.com>
Signed-off-by: NAlex Elder <aelder@sgi.com>

We set the IO_ISAIO flag for all read/write I/O since early Linux
2.6.x.  Remove it as it has lost it's purpose long ago.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <david@fromorbit.com>
Reviewed-by: NEric Sandeen <sandeen@sandeen.net>
Signed-off-by: NAlex Elder <aelder@sgi.com>

Summary of problem:

If a journal record wraps at the physical end of the journal, it has to be
read in two parts in xlog_do_recovery_pass(): a read at the physical end and a
read at the physical beginning.  If xlog_bread() has to re-align the first
read, the second read request does not take that re-alignment into account.
If the first read was re-aligned, the second read over-writes the end of the
data from the first read, effectively corrupting it.  This can happen either
when reading the record header or reading the record data.

The first sanity check in xlog_recover_process_data() is to check for a valid
clientid, so that is the error reported.

Summary of fix:

If there was a first read at the physical end, XFS_BUF_PTR() returns where the
data was requested to begin.  Conversely, because it is the result of
xlog_align(), offset indicates where the requested data for the first read
actually begins - whether or not xlog_bread() has re-aligned it.

Using offset as the base for the calculation of where to place the second read
data ensures that it will be correctly placed immediately following the data
from the first read instead of sometimes over-writing the end of it.

The attached patch has resolved the reported problem of occasional inability
to recover the journal (reporting "bad clientid").
Signed-off-by: NAndy Poling <andy@realbig.com>
Reviewed-by: NAlex Elder <aelder@sgi.com>
Signed-off-by: NAlex Elder <aelder@sgi.com>

Currently we have different end I/O handlers for read vs the different
types of write I/O.  But they are all very similar so we could just
use one with a few conditionals and reduce code size a lot.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NAlex Elder <aelder@sgi.com>
Signed-off-by: NAlex Elder <aelder@sgi.com>

The VM and I/O schedulers now expect us to use WRITE_SYNC_PLUG for
synchronous writeout.  Right now I can't see any changes in performance
numbers with this, but we're getting some beating for not using it,
and the knowledge definitely could help the block code to make better
decisions.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NAlex Elder <aelder@sgi.com>
Signed-off-by: NAlex Elder <aelder@sgi.com>

The iolock is used for protecting reads, writes and block truncates
against each other.  We have two classes of callers, the first one is
induced by a file operation and requires a reference to the inode be
held and not dropped after the operation is done:

 - xfs_vm_vmap, xfs_vn_fallocate, xfs_read, xfs_write, xfs_splice_read,
   xfs_splice_write and xfs_setattr are all implementations of VFS
   methods that require a live inode
 - xfs_getbmap and xfs_swap_extents are ioctl subcommand for which the
   same is true
 - xfs_truncate_file is only called on quota inodes just returned from
   xfs_iget
 - xfs_sync_inode_data does the lock just after an igrab()
 - xfs_filestream_associate and xfs_filestream_new_ag take the iolock
   on the parent inode of an inode which by VFS rules must be referenced

And we have various calls to truncate blocks past EOF or the whole
file when dropping the last reference to an inode.  Unfortunately
lockdep complains when we do memory allocations that can recurse into
the filesystem in the first class because the second class happens to
take the same lock.  To avoid this re-init the iolock in the beginning
of xfs_fs_clear_inode to get a new lock class.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NAlex Elder <aelder@sgi.com>
Signed-off-by: NAlex Elder <aelder@sgi.com>

When completing I/O requests we must not allow the memory allocator to
recurse into the filesystem, as we might deadlock on waiting for the
I/O completion otherwise.  The only thing currently allocating normal
GFP_KERNEL memory is the allocation of the transaction structure for
the unwritten extent conversion.  Add a memflags argument to
_xfs_trans_alloc to allow controlling the allocator behaviour.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reported-by: NThomas Neumann <tneumann@users.sourceforge.net>
Tested-by: NThomas Neumann <tneumann@users.sourceforge.net>
Reviewed-by: NAlex Elder <aelder@sgi.com>
Signed-off-by: NAlex Elder <aelder@sgi.com>