Some time ago, mkfs.xfs started picking the storage physical
sector size as the default filesystem "sector size" in order
to avoid RMW costs incurred by doing IOs at logical sector
size alignments.

However, this means that for a filesystem made with i.e.
a 4k sector size on an "advanced format" 4k/512 disk,
512-byte direct IOs are no longer allowed.  This means
that XFS has essentially turned this AF drive into a hard
4K device, from the filesystem on up.

XFS's mkfs-specified "sector size" is really just controlling
the minimum size & alignment of filesystem metadata.

There is no real need to tightly couple XFS's minimal
metadata size to the minimum allowed direct IO size;
XFS can continue doing metadata in optimal sizes, but
still allow smaller DIOs for apps which issue them,
for whatever reason.

This patch adds a new field to the xfs_buftarg, so that
we now track 2 sizes:

 1) The metadata sector size, which is the minimum unit and
    alignment of IO which will be performed by metadata operations.
 2) The device logical sector size

The first is used internally by the file system for metadata
alignment and IOs.
The second is used for the minimum allowed direct IO alignment.

This has passed xfstests on filesystems made with 4k sectors,
including when run under the patch I sent to ignore
XFS_IOC_DIOINFO, and issue 512 DIOs anyway.  I also directly
tested end of block behavior on preallocated, sparse, and
existing files when we do a 512 IO into a 4k file on a 
4k-sector filesystem, to be sure there were no unexpected
behaviors.
Signed-off-by: NEric Sandeen <sandeen@redhat.com>
Reviewed-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

In preparation for adding new members to the structure,
give these old ones more descriptive names:

	bt_ssize -> bt_meta_sectorsize
	bt_smask -> bt_meta_sectormask
Signed-off-by: NEric Sandeen <sandeen@redhat.com>
Reviewed-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Clean up the xfs_buftarg structure a bit:
- remove bt_bsize which is never used
- replace bt_sshift with bt_ssize; we only ever shift it back
Signed-off-by: NEric Sandeen <sandeen@redhat.com>
Reviewed-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

If we are doing aysnc writeback of metadata, we can get write errors
but have nobody to report them to. At the moment, we simply attempt
to reissue the write from io completion in the hope that it's a
transient error.

When it's not a transient error, the buffer is stuck forever in
this loop, and we cannot break out of it. Eventually, unmount will
hang because the AIL cannot be emptied and everything goes downhill
from them.

To solve this problem, only retry the write IO once before aborting
it. We don't throw the buffer away because some transient errors can
last minutes (e.g.  FC path failover) or even hours (thin
provisioned devices that have run out of backing space) before they
go away. Hence we really want to keep trying until we can't try any
more.

Because the buffer was not cleaned, however, it does not get removed
from the AIL and hence the next pass across the AIL will start IO on
it again. As such, we still get the "retry forever" semantics that
we currently have, but we allow other access to the buffer in the
mean time. Meanwhile the filesystem can continue to modify the
buffer and relog it, so the IO errors won't hang the log or the
filesystem.

Now when we are pushing the AIL, we can see all these "permanent IO
error" buffers and we can issue a warning about failures before we
retry the IO. We can also catch these buffers when unmounting an
issue a corruption warning, too.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NBen Myers <bpm@sgi.com>

The xfsbdstrat helper is a small but useless wrapper for xfs_buf_iorequest that
handles the case of a shut down filesystem.  Most of the users have private,
uncached buffers that can just be freed in this case, but the complex error
handling in xfs_bioerror_relse messes up the case when it's called without
a locked buffer.

Remove xfsbdstrat and opencode the error handling in the callers.  All but
one can simply return an error and don't need to deal with buffer state,
and the one caller that cares about the buffer state could do with a major
cleanup as well, but we'll defer that to later.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NBen Myers <bpm@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

The "verbose" argument to xfs_setsize_buftarg_flags() has been
unused since:

ffe37436 xfs: stop using the page cache to back the buffer cache

Remove it, and fold the function into xfs_setsize_buftarg()
now that there's no need for different types of callers.

Fix inconsistent comment spacing while we're at it.
Signed-off-by: NEric Sandeen <sandeen@redhat.com>
Reviewed-by: NBrian Foster <bfoster@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

xfs_trans.h has a dependency on xfs_log.h for a couple of
structures. Most code that does transactions doesn't need to know
anything about the log, but this dependency means that they have to
include xfs_log.h. Decouple the xfs_trans.h and xfs_log.h header
files and clean up the includes to be in dependency order.

In doing this, remove the direct include of xfs_trans_reserve.h from
xfs_trans.h so that we remove the dependency between xfs_trans.h and
xfs_mount.h. Hence the xfs_trans.h include can be moved to the
indicate the actual dependencies other header files have on it.

Note that these are kernel only header files, so this does not
translate to any userspace changes at all.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NBen Myers <bpm@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

__xfs_printk adds its own "\n".  Having it in the original string
leads to unintentional blank lines from these messages.

Most format strings have no newline, but a few do, leading to
i.e.:

[ 7347.119911] XFS (sdb2): Access to block zero in inode 132 start_block: 0 start_off: 0 blkcnt: 0 extent-state: 0 lastx: 1a05
[ 7347.119911] 
[ 7347.119919] XFS (sdb2): Access to block zero in inode 132 start_block: 0 start_off: 0 blkcnt: 0 extent-state: 0 lastx: 1a05
[ 7347.119919] 

Fix them all.
Signed-off-by: NEric Sandeen <sandeen@redhat.com>
Reviewed-by: NMark Tinguely <tinguely@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

This patch adds the missing call to list_lru_destroy (spotted by Li Zhong)
and moves the deletion to after the shrinker is unregistered, as correctly
spotted by Dave
Signed-off-by: NGlauber Costa <glommer@openvz.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Dave Chinner <dchinner@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

We currently use a compile-time constant to size the node array for the
list_lru structure.  Due to this, we don't need to allocate any memory at
initialization time.  But as a consequence, the structures that contain
embedded list_lru lists can become way too big (the superblock for
instance contains two of them).

This patch aims at ameliorating this situation by dynamically allocating
the node arrays with the firmware provided nr_node_ids.
Signed-off-by: NGlauber Costa <glommer@openvz.org>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Cc: Arve Hjønnevåg <arve@android.com>
Cc: Carlos Maiolino <cmaiolino@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Chuck Lever <chuck.lever@oracle.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: David Rientjes <rientjes@google.com>
Cc: Gleb Natapov <gleb@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: J. Bruce Fields <bfields@redhat.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: John Stultz <john.stultz@linaro.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Kent Overstreet <koverstreet@google.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Cc: Thomas Hellstrom <thellstrom@vmware.com>
Cc: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

In converting the buffer lru lists to use the generic code, the locking
for marking the buffers as on the dispose list was lost.  This results in
confusion in LRU buffer tracking and acocunting, resulting in reference
counts being mucked up and filesystem beig unmountable.

To fix this, introduce an internal buffer spinlock to protect the state
field that holds the dispose list information.  Because there is now
locking needed around xfs_buf_lru_add/del, and they are used in exactly
one place each two lines apart, get rid of the wrappers and code the logic
directly in place.

Further, the LRU emptying code used on unmount is less than optimal.
Convert it to use a dispose list as per a normal shrinker walk, and repeat
the walk that fills the dispose list until the LRU is empty.  Thi avoids
needing to drop and regain the LRU lock for every item being freed, and
allows the same logic as the shrinker isolate call to be used.  Simpler,
easier to understand.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NGlauber Costa <glommer@openvz.org>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Cc: Arve Hjønnevåg <arve@android.com>
Cc: Carlos Maiolino <cmaiolino@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Chuck Lever <chuck.lever@oracle.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: David Rientjes <rientjes@google.com>
Cc: Gleb Natapov <gleb@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: J. Bruce Fields <bfields@redhat.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: John Stultz <john.stultz@linaro.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Kent Overstreet <koverstreet@google.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Cc: Thomas Hellstrom <thellstrom@vmware.com>
Cc: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

fix warnings

Cc: Dave Chinner <dchinner@redhat.com>
Cc: Glauber Costa <glommer@openvz.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Convert the buftarg LRU to use the new generic LRU list and take advantage
of the functionality it supplies to make the buffer cache shrinker node
aware.
Signed-off-by: NGlauber Costa <glommer@openvz.org>
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Cc: Arve Hjønnevåg <arve@android.com>
Cc: Carlos Maiolino <cmaiolino@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Chuck Lever <chuck.lever@oracle.com>
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: David Rientjes <rientjes@google.com>
Cc: Gleb Natapov <gleb@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: J. Bruce Fields <bfields@redhat.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: John Stultz <john.stultz@linaro.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Kent Overstreet <koverstreet@google.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Cc: Thomas Hellstrom <thellstrom@vmware.com>
Cc: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Signed-off-by: NZhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Reviewed-by: NMark Tinguely <tinguely@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Signed-off-by: NZhi Yong Wu <wuzhy@linux.vnet.ibm.com>
Reviewed-by: NMark Tinguely <tinguely@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

The transaction reservation size calculations is used by both kernel
and userspace, but most of the transaction code in xfs_trans.c is
kernel specific. Split all the transaction reservation code out into
it's own files to make sharing with userspace simpler. This just
leaves kernel-only definitions in xfs_trans.h, so it doesn't need to
be shared with userspace anymore, either.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NMark Tinguely <tinguely@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Note: this changes the on-disk remote attribute format. I assert
that this is OK to do as CRCs are marked experimental and the first
kernel it is included in has not yet reached release yet. Further,
the userspace utilities are still evolving and so anyone using this
stuff right now is a developer or tester using volatile filesystems
for testing this feature. Hence changing the format right now to
save longer term pain is the right thing to do.

The fundamental change is to move from a header per extent in the
attribute to a header per filesytem block in the attribute. This
means there are more header blocks and the parsing of the attribute
data is slightly more complex, but it has the advantage that we
always know the size of the attribute on disk based on the length of
the data it contains.

This is where the header-per-extent method has problems. We don't
know the size of the attribute on disk without first knowing how
many extents are used to hold it. And we can't tell from a
mapping lookup, either, because remote attributes can be allocated
contiguously with other attribute blocks and so there is no obvious
way of determining the actual size of the atribute on disk short of
walking and mapping buffers.

The problem with this approach is that if we map a buffer
incorrectly (e.g. we make the last buffer for the attribute data too
long), we then get buffer cache lookup failure when we map it
correctly. i.e. we get a size mismatch on lookup. This is not
necessarily fatal, but it's a cache coherency problem that can lead
to returning the wrong data to userspace or writing the wrong data
to disk. And debug kernels will assert fail if this occurs.

I found lots of niggly little problems trying to fix this issue on a
4k block size filesystem, finally getting it to pass with lots of
fixes. The thing is, 1024 byte filesystems still failed, and it was
getting really complex handling all the corner cases that were
showing up. And there were clearly more that I hadn't found yet.

It is complex, fragile code, and if we don't fix it now, it will be
complex, fragile code forever more.

Hence the simple fix is to add a header to each filesystem block.
This gives us the same relationship between the attribute data
length and the number of blocks on disk as we have without CRCs -
it's a linear mapping and doesn't require us to guess anything. It
is simple to implement, too - the remote block count calculated at
lookup time can be used by the remote attribute set/get/remove code
without modification for both CRC and non-CRC filesystems. The world
becomes sane again.

Because the copy-in and copy-out now need to iterate over each
filesystem block, I moved them into helper functions so we separate
the block mapping and buffer manupulations from the attribute data
and CRC header manipulations. The code becomes much clearer as a
result, and it is a lot easier to understand and debug. It also
appears to be much more robust - once it worked on 4k block size
filesystems, it has worked without failure on 1k block size
filesystems, too.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NBen Myers <bpm@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

(cherry picked from commit ad1858d7)

There are several places where we use KM_SLEEP allocation contexts
and use the fact that they are called from transaction context to
add KM_NOFS where appropriate. Unfortunately, there are several
places where the code makes this assumption but can be called from
outside transaction context but with filesystem locks held. These
places need explicit KM_NOFS annotations to avoid lockdep
complaining about reclaim contexts.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NBen Myers <bpm@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

(cherry picked from commit ac14876c)

Note: this changes the on-disk remote attribute format. I assert
that this is OK to do as CRCs are marked experimental and the first
kernel it is included in has not yet reached release yet. Further,
the userspace utilities are still evolving and so anyone using this
stuff right now is a developer or tester using volatile filesystems
for testing this feature. Hence changing the format right now to
save longer term pain is the right thing to do.

The fundamental change is to move from a header per extent in the
attribute to a header per filesytem block in the attribute. This
means there are more header blocks and the parsing of the attribute
data is slightly more complex, but it has the advantage that we
always know the size of the attribute on disk based on the length of
the data it contains.

This is where the header-per-extent method has problems. We don't
know the size of the attribute on disk without first knowing how
many extents are used to hold it. And we can't tell from a
mapping lookup, either, because remote attributes can be allocated
contiguously with other attribute blocks and so there is no obvious
way of determining the actual size of the atribute on disk short of
walking and mapping buffers.

The problem with this approach is that if we map a buffer
incorrectly (e.g. we make the last buffer for the attribute data too
long), we then get buffer cache lookup failure when we map it
correctly. i.e. we get a size mismatch on lookup. This is not
necessarily fatal, but it's a cache coherency problem that can lead
to returning the wrong data to userspace or writing the wrong data
to disk. And debug kernels will assert fail if this occurs.

I found lots of niggly little problems trying to fix this issue on a
4k block size filesystem, finally getting it to pass with lots of
fixes. The thing is, 1024 byte filesystems still failed, and it was
getting really complex handling all the corner cases that were
showing up. And there were clearly more that I hadn't found yet.

It is complex, fragile code, and if we don't fix it now, it will be
complex, fragile code forever more.

Hence the simple fix is to add a header to each filesystem block.
This gives us the same relationship between the attribute data
length and the number of blocks on disk as we have without CRCs -
it's a linear mapping and doesn't require us to guess anything. It
is simple to implement, too - the remote block count calculated at
lookup time can be used by the remote attribute set/get/remove code
without modification for both CRC and non-CRC filesystems. The world
becomes sane again.

Because the copy-in and copy-out now need to iterate over each
filesystem block, I moved them into helper functions so we separate
the block mapping and buffer manupulations from the attribute data
and CRC header manipulations. The code becomes much clearer as a
result, and it is a lot easier to understand and debug. It also
appears to be much more robust - once it worked on 4k block size
filesystems, it has worked without failure on 1k block size
filesystems, too.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NBen Myers <bpm@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

There are several places where we use KM_SLEEP allocation contexts
and use the fact that they are called from transaction context to
add KM_NOFS where appropriate. Unfortunately, there are several
places where the code makes this assumption but can be called from
outside transaction context but with filesystem locks held. These
places need explicit KM_NOFS annotations to avoid lockdep
complaining about reclaim contexts.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NBen Myers <bpm@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Failed buffer readahead can leave the buffer in the cache marked
with an error. Most callers that then issue a subsequent read on the
buffer do not zero the b_error field out, and so we may incorectly
detect an error during IO completion due to the stale error value
left on the buffer.

Avoid this problem by zeroing the error before IO submission. This
ensures that the only IO errors that are detected those captured
from are those captured from bio submission or completion.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NMark Tinguely <tinguely@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

(cherry picked from commit c163f9a1)

Failed buffer readahead can leave the buffer in the cache marked
with an error. Most callers that then issue a subsequent read on the
buffer do not zero the b_error field out, and so we may incorectly
detect an error during IO completion due to the stale error value
left on the buffer.

Avoid this problem by zeroing the error before IO submission. This
ensures that the only IO errors that are detected those captured
from are those captured from bio submission or completion.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NMark Tinguely <tinguely@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

When we read a buffer, we might get an error from the underlying
block device and not the real data. Hence if we get an IO error, we
shouldn't run the verifier but instead just pass the IO error
straight through.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NMark Tinguely <tinguely@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

The trylock log force invoked via xfs_buf_item_push() can attempt
to acquire xa_lock, thus leading to a recursion bug when called
with xa_lock held.

This log force was originally added to xfs_buf_trylock() to address
xfsaild stalls due to pinned and stale buffers. Since the addition
of this behavior, the log item pushing code had been reworked to
detect and track pinned items to inform xfsaild to issue a log
force itself when necessary. As such, the log force on trylock
failure is redundant and safe to remove.
Signed-off-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

When the new inode verify in xfs_iread() fails, the create
transaction is aborted and a shutdown occurs. The subsequent unmount
then hangs in xfs_wait_buftarg() on a buffer that has an elevated
hold count. Debug showed that it was an AGI buffer getting stuck:

[   22.576147] XFS (vdb): buffer 0x2/0x1, hold 0x2 stuck
[   22.976213] XFS (vdb): buffer 0x2/0x1, hold 0x2 stuck
[   23.376206] XFS (vdb): buffer 0x2/0x1, hold 0x2 stuck
[   23.776325] XFS (vdb): buffer 0x2/0x1, hold 0x2 stuck

The trace of this buffer leading up to the shutdown (trimmed for
brevity) looks like:

xfs_buf_init:        bno 0x2 nblks 0x1 hold 1 caller xfs_buf_get_map
xfs_buf_get:         bno 0x2 len 0x200 hold 1 caller xfs_buf_read_map
xfs_buf_read:        bno 0x2 len 0x200 hold 1 caller xfs_trans_read_buf_map
xfs_buf_iorequest:   bno 0x2 nblks 0x1 hold 1 caller _xfs_buf_read
xfs_buf_hold:        bno 0x2 nblks 0x1 hold 1 caller xfs_buf_iorequest
xfs_buf_rele:        bno 0x2 nblks 0x1 hold 2 caller xfs_buf_iorequest
xfs_buf_iowait:      bno 0x2 nblks 0x1 hold 1 caller _xfs_buf_read
xfs_buf_ioerror:     bno 0x2 len 0x200 hold 1 caller xfs_buf_bio_end_io
xfs_buf_iodone:      bno 0x2 nblks 0x1 hold 1 caller _xfs_buf_ioend
xfs_buf_iowait_done: bno 0x2 nblks 0x1 hold 1 caller _xfs_buf_read
xfs_buf_hold:        bno 0x2 nblks 0x1 hold 1 caller xfs_buf_item_init
xfs_trans_read_buf:  bno 0x2 len 0x200 hold 2 recur 0 refcount 1
xfs_trans_brelse:    bno 0x2 len 0x200 hold 2 recur 0 refcount 1
xfs_buf_item_relse:  bno 0x2 nblks 0x1 hold 2 caller xfs_trans_brelse
xfs_buf_rele:        bno 0x2 nblks 0x1 hold 2 caller xfs_buf_item_relse
xfs_buf_unlock:      bno 0x2 nblks 0x1 hold 1 caller xfs_trans_brelse
xfs_buf_rele:        bno 0x2 nblks 0x1 hold 1 caller xfs_trans_brelse
xfs_buf_trylock:     bno 0x2 nblks 0x1 hold 2 caller _xfs_buf_find
xfs_buf_find:        bno 0x2 len 0x200 hold 2 caller xfs_buf_get_map
xfs_buf_get:         bno 0x2 len 0x200 hold 2 caller xfs_buf_read_map
xfs_buf_read:        bno 0x2 len 0x200 hold 2 caller xfs_trans_read_buf_map
xfs_buf_hold:        bno 0x2 nblks 0x1 hold 2 caller xfs_buf_item_init
xfs_trans_read_buf:  bno 0x2 len 0x200 hold 3 recur 0 refcount 1
xfs_trans_log_buf:   bno 0x2 len 0x200 hold 3 recur 0 refcount 1
xfs_buf_item_unlock: bno 0x2 len 0x200 hold 3 flags DIRTY liflags ABORTED
xfs_buf_unlock:      bno 0x2 nblks 0x1 hold 3 caller xfs_buf_item_unlock
xfs_buf_rele:        bno 0x2 nblks 0x1 hold 3 caller xfs_buf_item_unlock

And that is the AGI buffer from cold cache read into memory to
transaction abort. You can see at transaction abort the bli is dirty
and only has a single reference. The item is not pinned, and it's
not in the AIL. Hence the only reference to it is this transaction.

The problem is that the xfs_buf_item_unlock() call is dropping the
last reference to the xfs_buf_log_item attached to the buffer (which
holds a reference to the buffer), but it is not freeing the
xfs_buf_log_item. Hence nothing will ever release the buffer, and
the unmount hangs waiting for this reference to go away.

The fix is simple - xfs_buf_item_unlock needs to detect the last
reference going away in this case and free the xfs_buf_log_item to
release the reference it holds on the buffer.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NBen Myers <bpm@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

When _xfs_buf_find is passed an out of range address, it will fail
to find a relevant struct xfs_perag and oops with a null
dereference. This can happen when trying to walk a filesystem with a
metadata inode that has a partially corrupted extent map (i.e. the
block number returned is corrupt, but is otherwise intact) and we
try to read from the corrupted block address.

In this case, just fail the lookup. If it is readahead being issued,
it will simply not be done, but if it is real read that fails we
will get an error being reported.  Ideally this case should result
in an EFSCORRUPTED error being reported, but we cannot return an
error through xfs_buf_read() or xfs_buf_get() so this lookup failure
may result in ENOMEM or EIO errors being reported instead.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NBen Myers <bpm@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

When the new inode verify in xfs_iread() fails, the create
transaction is aborted and a shutdown occurs. The subsequent unmount
then hangs in xfs_wait_buftarg() on a buffer that has an elevated
hold count. Debug showed that it was an AGI buffer getting stuck:

[   22.576147] XFS (vdb): buffer 0x2/0x1, hold 0x2 stuck
[   22.976213] XFS (vdb): buffer 0x2/0x1, hold 0x2 stuck
[   23.376206] XFS (vdb): buffer 0x2/0x1, hold 0x2 stuck
[   23.776325] XFS (vdb): buffer 0x2/0x1, hold 0x2 stuck

The trace of this buffer leading up to the shutdown (trimmed for
brevity) looks like:

xfs_buf_init:        bno 0x2 nblks 0x1 hold 1 caller xfs_buf_get_map
xfs_buf_get:         bno 0x2 len 0x200 hold 1 caller xfs_buf_read_map
xfs_buf_read:        bno 0x2 len 0x200 hold 1 caller xfs_trans_read_buf_map
xfs_buf_iorequest:   bno 0x2 nblks 0x1 hold 1 caller _xfs_buf_read
xfs_buf_hold:        bno 0x2 nblks 0x1 hold 1 caller xfs_buf_iorequest
xfs_buf_rele:        bno 0x2 nblks 0x1 hold 2 caller xfs_buf_iorequest
xfs_buf_iowait:      bno 0x2 nblks 0x1 hold 1 caller _xfs_buf_read
xfs_buf_ioerror:     bno 0x2 len 0x200 hold 1 caller xfs_buf_bio_end_io
xfs_buf_iodone:      bno 0x2 nblks 0x1 hold 1 caller _xfs_buf_ioend
xfs_buf_iowait_done: bno 0x2 nblks 0x1 hold 1 caller _xfs_buf_read
xfs_buf_hold:        bno 0x2 nblks 0x1 hold 1 caller xfs_buf_item_init
xfs_trans_read_buf:  bno 0x2 len 0x200 hold 2 recur 0 refcount 1
xfs_trans_brelse:    bno 0x2 len 0x200 hold 2 recur 0 refcount 1
xfs_buf_item_relse:  bno 0x2 nblks 0x1 hold 2 caller xfs_trans_brelse
xfs_buf_rele:        bno 0x2 nblks 0x1 hold 2 caller xfs_buf_item_relse
xfs_buf_unlock:      bno 0x2 nblks 0x1 hold 1 caller xfs_trans_brelse
xfs_buf_rele:        bno 0x2 nblks 0x1 hold 1 caller xfs_trans_brelse
xfs_buf_trylock:     bno 0x2 nblks 0x1 hold 2 caller _xfs_buf_find
xfs_buf_find:        bno 0x2 len 0x200 hold 2 caller xfs_buf_get_map
xfs_buf_get:         bno 0x2 len 0x200 hold 2 caller xfs_buf_read_map
xfs_buf_read:        bno 0x2 len 0x200 hold 2 caller xfs_trans_read_buf_map
xfs_buf_hold:        bno 0x2 nblks 0x1 hold 2 caller xfs_buf_item_init
xfs_trans_read_buf:  bno 0x2 len 0x200 hold 3 recur 0 refcount 1
xfs_trans_log_buf:   bno 0x2 len 0x200 hold 3 recur 0 refcount 1
xfs_buf_item_unlock: bno 0x2 len 0x200 hold 3 flags DIRTY liflags ABORTED
xfs_buf_unlock:      bno 0x2 nblks 0x1 hold 3 caller xfs_buf_item_unlock
xfs_buf_rele:        bno 0x2 nblks 0x1 hold 3 caller xfs_buf_item_unlock

And that is the AGI buffer from cold cache read into memory to
transaction abort. You can see at transaction abort the bli is dirty
and only has a single reference. The item is not pinned, and it's
not in the AIL. Hence the only reference to it is this transaction.

The problem is that the xfs_buf_item_unlock() call is dropping the
last reference to the xfs_buf_log_item attached to the buffer (which
holds a reference to the buffer), but it is not freeing the
xfs_buf_log_item. Hence nothing will ever release the buffer, and
the unmount hangs waiting for this reference to go away.

The fix is simple - xfs_buf_item_unlock needs to detect the last
reference going away in this case and free the xfs_buf_log_item to
release the reference it holds on the buffer.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NBen Myers <bpm@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

When _xfs_buf_find is passed an out of range address, it will fail
to find a relevant struct xfs_perag and oops with a null
dereference. This can happen when trying to walk a filesystem with a
metadata inode that has a partially corrupted extent map (i.e. the
block number returned is corrupt, but is otherwise intact) and we
try to read from the corrupted block address.

In this case, just fail the lookup. If it is readahead being issued,
it will simply not be done, but if it is real read that fails we
will get an error being reported.  Ideally this case should result
in an EFSCORRUPTED error being reported, but we cannot return an
error through xfs_buf_read() or xfs_buf_get() so this lookup failure
may result in ENOMEM or EIO errors being reported instead.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NBrian Foster <bfoster@redhat.com>
Reviewed-by: NBen Myers <bpm@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Commits starting at 77c1a08f introduced a multiple segment support
to xfs_buf. xfs_trans_buf_item_match() could not find a multi-segment
buffer in the transaction because it was looking at the single segment
block number rather than the multi-segment b_maps[0].bm.bn. This
results on a recursive buffer lock that can never be satisfied.

This patch:
 1) Changed the remaining b_map accesses to be b_maps[0] accesses.
 2) Renames the single segment b_map structure to __b_map to avoid
    future confusion.
Signed-off-by: NMark Tinguely <tinguely@sgi.com>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NBen Myers <bpm@sgi.com>

Commits starting at 77c1a08f introduced a multiple segment support
to xfs_buf. xfs_trans_buf_item_match() could not find a multi-segment
buffer in the transaction because it was looking at the single segment
block number rather than the multi-segment b_maps[0].bm.bn. This
results on a recursive buffer lock that can never be satisfied.

This patch:
 1) Changed the remaining b_map accesses to be b_maps[0] accesses.
 2) Renames the single segment b_map structure to __b_map to avoid
    future confusion.
Signed-off-by: NMark Tinguely <tinguely@sgi.com>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NBen Myers <bpm@sgi.com>

Error handling in xfs_buf_ioapply_map() does not handle IO reference
counts correctly. We increment the b_io_remaining count before
building the bio, but then fail to decrement it in the failure case.
This leads to the buffer never running IO completion and releasing
the reference that the IO holds, so at unmount we can leak the
buffer. This leak is captured by this assert failure during unmount:

XFS: Assertion failed: atomic_read(&pag->pag_ref) == 0, file: fs/xfs/xfs_mount.c, line: 273

This is not a new bug - the b_io_remaining accounting has had this
problem for a long, long time - it's just very hard to get a
zero length bio being built by this code...

Further, the buffer IO error can be overwritten on a multi-segment
buffer by subsequent bio completions for partial sections of the
buffer. Hence we should only set the buffer error status if the
buffer is not already carrying an error status. This ensures that a
partial IO error on a multi-segment buffer will not be lost. This
part of the problem is a regression, however.

cc: <stable@vger.kernel.org>
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NMark Tinguely <tinguely@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

To separate the verifiers from iodone functions and associate read
and write verifiers at the same time, introduce a buffer verifier
operations structure to the xfs_buf.

This avoids the need for assigning the write verifier, clearing the
iodone function and re-running ioend processing in the read
verifier, and gets rid of the nasty "b_pre_io" name for the write
verifier function pointer. If we ever need to, it will also be
easier to add further content specific callbacks to a buffer with an
ops structure in place.

We also avoid needing to export verifier functions, instead we
can simply export the ops structures for those that are needed
outside the function they are defined in.

This patch also fixes a directory block readahead verifier issue
it exposed.

This patch also adds ops callbacks to the inode/alloc btree blocks
initialised by growfs. These will need more work before they will
work with CRCs.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NPhil White <pwhite@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Add a callback to the buffer write path to enable verification of
the buffer and CRC calculation prior to issuing the write to the
underlying storage.

If the callback function detects some kind of failure or error
condition, it must mark the buffer with an error so that the caller
can take appropriate action. In the case of xfs_buf_ioapply(), a
corrupt metadta buffer willt rigger a shutdown of the filesystem,
because something is clearly wrong and we can't allow corrupt
metadata to be written to disk.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NPhil White <pwhite@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

With verification being done as an IO completion callback, different
errors can be returned from a read. Uncached reads only return a
buffer or NULL on failure, which means the verification error cannot
be returned to the caller.

Split the error handling for these reads into two - a failure to get
a buffer will still return NULL, but a read error will return a
referenced buffer with b_error set rather than NULL. The caller is
responsible for checking the error state of the buffer returned.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NPhil White <pwhite@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Add a verifier function callback capability to the buffer read
interfaces.  This will be used by the callers to supply a function
that verifies the contents of the buffer when it is read from disk.
This patch does not provide callback functions, but simply modifies
the interfaces to allow them to be called.

The reason for adding this to the read interfaces is that it is very
difficult to tell fom the outside is a buffer was just read from
disk or whether we just pulled it out of cache. Supplying a callbck
allows the buffer cache to use it's internal knowledge of the buffer
to execute it only when the buffer is read from disk.

It is intended that the verifier functions will mark the buffer with
an EFSCORRUPTED error when verification fails. This allows the
reading context to distinguish a verification error from an IO
error, and potentially take further actions on the buffer (e.g.
attempt repair) based on the error reported.
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NPhil White <pwhite@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

Error handling in xfs_buf_ioapply_map() does not handle IO reference
counts correctly. We increment the b_io_remaining count before
building the bio, but then fail to decrement it in the failure case.
This leads to the buffer never running IO completion and releasing
the reference that the IO holds, so at unmount we can leak the
buffer. This leak is captured by this assert failure during unmount:

XFS: Assertion failed: atomic_read(&pag->pag_ref) == 0, file: fs/xfs/xfs_mount.c, line: 273

This is not a new bug - the b_io_remaining accounting has had this
problem for a long, long time - it's just very hard to get a
zero length bio being built by this code...

Further, the buffer IO error can be overwritten on a multi-segment
buffer by subsequent bio completions for partial sections of the
buffer. Hence we should only set the buffer error status if the
buffer is not already carrying an error status. This ensures that a
partial IO error on a multi-segment buffer will not be lost. This
part of the problem is a regression, however.

cc: <stable@vger.kernel.org>
Signed-off-by: NDave Chinner <dchinner@redhat.com>
Reviewed-by: NMark Tinguely <tinguely@sgi.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

While xfs_buftarg_shrink() is freeing buffers from the dispose list (filled with
buffers from lru list), there is a possibility to have xfs_buf_stale() racing
with it, and removing buffers from dispose list before xfs_buftarg_shrink() does
it.

This happens because xfs_buftarg_shrink() handle the dispose list without
locking and the test condition in xfs_buf_stale() checks for the buffer being in
*any* list:

if (!list_empty(&bp->b_lru))

If the buffer happens to be on dispose list, this causes the buffer counter of
lru list (btp->bt_lru_nr) to be decremented twice (once in xfs_buftarg_shrink()
and another in xfs_buf_stale()) causing a wrong account usage of the lru list.

This may cause xfs_buftarg_shrink() to return a wrong value to the memory
shrinker shrink_slab(), and such account error may also cause an underflowed
value to be returned; since the counter is lower than the current number of
items in the lru list, a decrement may happen when the counter is 0, causing
an underflow on the counter.

The fix uses a new flag field (and a new buffer flag) to serialize buffer
handling during the shrink process. The new flag field has been designed to use
btp->bt_lru_lock/unlock instead of xfs_buf_lock/unlock mechanism.

dchinner, sandeen, aquini and aris also deserve credits for this.
Signed-off-by: NCarlos Maiolino <cmaiolino@redhat.com>
Reviewed-by: NBen Myers <bpm@sgi.com>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

While xfs_buftarg_shrink() is freeing buffers from the dispose list (filled with
buffers from lru list), there is a possibility to have xfs_buf_stale() racing
with it, and removing buffers from dispose list before xfs_buftarg_shrink() does
it.

This happens because xfs_buftarg_shrink() handle the dispose list without
locking and the test condition in xfs_buf_stale() checks for the buffer being in
*any* list:

if (!list_empty(&bp->b_lru))

If the buffer happens to be on dispose list, this causes the buffer counter of
lru list (btp->bt_lru_nr) to be decremented twice (once in xfs_buftarg_shrink()
and another in xfs_buf_stale()) causing a wrong account usage of the lru list.

This may cause xfs_buftarg_shrink() to return a wrong value to the memory
shrinker shrink_slab(), and such account error may also cause an underflowed
value to be returned; since the counter is lower than the current number of
items in the lru list, a decrement may happen when the counter is 0, causing
an underflow on the counter.

The fix uses a new flag field (and a new buffer flag) to serialize buffer
handling during the shrink process. The new flag field has been designed to use
btp->bt_lru_lock/unlock instead of xfs_buf_lock/unlock mechanism.

dchinner, sandeen, aquini and aris also deserve credits for this.
Signed-off-by: NCarlos Maiolino <cmaiolino@redhat.com>
Reviewed-by: NBen Myers <bpm@sgi.com>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

xfs_bdstrat_cb only adds a check for a shutdown filesystem over
xfs_buf_iorequest, but xfs_buf_iodone_callbacks just checked for a shut down
filesystem a little earlier.  In addition the shutdown handling in
xfs_bdstrat_cb is not very suitable for this caller.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>

If the b_iodone handler is run in calling context in xfs_buf_iorequest we
can run into a recursion where xfs_buf_iodone_callbacks keeps calling back
into xfs_buf_iorequest because an I/O error happened, which keeps calling
back into xfs_buf_iorequest.  This chain will usually not take long
because the filesystem gets shut down because of log I/O errors, but even
over a short time it can cause stack overflows if run on the same context.

As a short term workaround make sure we always call the iodone handler in
workqueue context.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NDave Chinner <dchinner@redhat.com>
Signed-off-by: NBen Myers <bpm@sgi.com>