When free space is running low the extent allocator may choose to allocate
an extent from an AG without leaving sufficient space for a btree split
when inserting the new extent (see where xfs_bmap_btalloc() sets minleft
to 0). In this case the allocator will enable the lowspace algorithm which
is supposed to allow further allocations (such as btree splits and
newroots) to allocate from sequential AGs. This algorithm has been broken
for a long time and this patch restores its behaviour.

SGI-PV: 983338

SGI-Modid: xfs-linux-melb:xfs-kern:31358a
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NDavid Chinner <dgc@sgi.com>

The bmap btree split code relies on a previous data extent allocation
(from xfs_bmap_btalloc()) to find an AG that has sufficient space to
perform a full btree split, when inserting the extent. When converting
unwritten extents we don't allocate a data extent so a btree split will be
the first allocation. In this case we need to set minleft so the allocator
will pick an AG that has space to complete the split(s).

SGI-PV: 983338

SGI-Modid: xfs-linux-melb:xfs-kern:31357a
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NDavid Chinner <dgc@sgi.com>

xfs_attrmulti_by_handle currently request the size based on
sizeof(attr_multiop_t) but should be using sizeof(xfs_attr_multiop_t)
because that is what it is dealing with. Despite beeing wrong this
actually harmless in practice because both structures are the same size on
all platforms.

But this sizeof was the only user of struct attr_multiop so we can just
kill it. Also move the ATTR_OP_* defines xfs_attr.h into the struct
xfs_attr_multiop defintion in xfs_fs.h because they are only used with
that structure, and are part of the user ABI for the
XFS_IOC_ATTRMULTI_BY_HANDLE ioctl.

SGI-PV: 983508

SGI-Modid: xfs-linux-melb:xfs-kern:31352a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NTim Shimmin <tes@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

xfs_attrlist_by_handle should only take the ATTR_ flags for the root
namespaces. The ATTR_KERN* flags may change at anytime and expect special
preconditions that can't be guaranteed for userspace-originating requests.
For example passing down ATTR_KERNNOVAL through xfs_attrlist_by_handle
will hit an assert in debug builds currently.

SGI-PV: 983677

SGI-Modid: xfs-linux-melb:xfs-kern:31351a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NTim Shimmin <tes@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

Instead of comparing buffer pointers, compare buffer block numbers and
don't keep buff

SGI-PV: 983564

SGI-Modid: xfs-linux-melb:xfs-kern:31346a
Signed-off-by: NBarry Naujok <bnaujok@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

Add missing file fs/xfs/linux-2.6/xfs_xattr.c

SGI-PV: 982343

SGI-Modid: xfs-linux-melb:xfs-kern:31234a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NTim Shimmin <tes@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

After a btree insert operation a cursor can be invalid due to block splits
and a maybe a new root block. We reset the cursor in xfs_bmbt_insert() in
the cases where we think we need to but it isn't enough as we still see
assertions. Just do what we do elsewhere and reset the cursor
unconditionally. Also remove the fix to revalidate the original cursor in
xfs_bmbt_insert().

SGI-PV: 983336

SGI-Modid: xfs-linux-melb:xfs-kern:31342a
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NDavid Chinner <dgc@sgi.com>

ASSERTs are no good to us on a non-debug build so use
XFS_WANT_CORRUPTED_GOTOs to report extent btree corruption ASAP.

SGI-PV: 983500

SGI-Modid: xfs-linux-melb:xfs-kern:31338a
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

xfs_dir2_node_lookup() calls xfs_da_node_lookup_int() which iterates
through leaf blocks containing the matching hash value for the name being
looked up. Inside xfs_da_node_lookup_int(), it calls the
xfs_dir2_leafn_lookup_for_entry() for each leaf block.
xfs_dir2_leafn_lookup_for_entry() iterates through each matching
hash/offset pair doing a name comparison to find the matching dirent.

For CI mode, the state->extrablk retains the details of the block that has
the CI match so xfs_dir2_node_lookup() can return the case-preserved name.

The original implementation didn't retain the xfs_da_buf_t properly, so
the lookup was returning a bogus name to be stored in the dentry.

In the case of unlink, the bad name was passed and in debug mode, ASSERTed
when it can't find the entry.

SGI-PV: 983284

SGI-Modid: xfs-linux-melb:xfs-kern:31337a
Signed-off-by: NBarry Naujok <bnaujok@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

The core kernel uses vfs_getattr to look at the inode size and similar
attributes, so there is no need to keep i_size uptodate for directories or
special files. This means we can remove xfs_validate_fields because the
I/O path already keeps i_size uptodate for regular files.

SGI-PV: 981498

SGI-Modid: xfs-linux-melb:xfs-kern:31336a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NBarry Naujok <bnaujok@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

xfs_remove and xfs_rmdir are almost the same with a little more work
performed in xfs_rmdir due to the . and .. entries. This patch merges
xfs_rmdir into xfs_remove and performs these actions conditionally.

Also clean up the error handling which was a nightmare in both versions
before.

SGI-PV: 981498

SGI-Modid: xfs-linux-melb:xfs-kern:31335a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NBarry Naujok <bnaujok@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

context count of ssize_t versus int. Change context count to be ssize_t.

SGI-PV: 983395

SGI-Modid: xfs-linux-melb:xfs-kern:31333a
Signed-off-by: NTim Shimmin <tes@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

This function is used to compact the indirect extent list by moving
extents from one page to the previous to fill them up. After we move some
extents to an earlier page we need to shuffle the remaining extents to the
start of the page. The actual bug here is the second argument to memmove()
needs to index past the extents, that were copied to the previous page,
and move the remaining extents. For pages that are already full (ie
ext_avail == 0) the compaction code has no net effect so don't do it.

SGI-PV: 983337

SGI-Modid: xfs-linux-melb:xfs-kern:31332a
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

During a forced shutdown a xfs inode can be destroyed before log I/O
involving that inode is complete. We need to wait for the inode to be
unpinned before tearing it down. Version 2 cleans up the code a bit by
relying on xfs_iflush() to do the unpinning and forced shutdown check.

SGI-PV: 981240

SGI-Modid: xfs-linux-melb:xfs-kern:31326a
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>
Signed-off-by: NDavid Chinner <dgc@sgi.com>

and not go through xfs_attr_list.

SGI-PV: 983395

SGI-Modid: xfs-linux-melb:xfs-kern:31324a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NTim Shimmin <tes@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

SGI-PV: 983394

SGI-Modid: xfs-linux-melb:xfs-kern:31323a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NTim Shimmin <tes@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

architecture.

This should fix the longstanding issues with xfs and old ABI arm boxes,
which lead to various asserts and xfs shutdowns, and for which an
(incorrect) patch has been floating around for years.

I've verified this patch by comparing the on-disk structure layouts using
pahole from the dwarves package, as well as running through a bit of xfsqa
under qemu-arm, modified so that the check/repair phase after each test
actually executes check/repair from the x86 host, on the filesystem
populated by the arm emulator. Thus far it all looks good.

There are 2 other structures with extra padding at the end, but they don't
seem to cause trouble. I suppose they could be packed as well:
xfs_dir2_data_unused_t and xfs_dir2_sf_t.

Note that userspace needs a similar treatment, and any filesystems which
were running with the previous rogue "fix" will now see corruption (either
in the kernel, or during xfs_repair) with this fix properly in place; it
may be worth teaching xfs_repair to identify and fix that specific issue.

SGI-PV: 982930

SGI-Modid: xfs-linux-melb:xfs-kern:31280a
Signed-off-by: NEric Sandeen <sandeen@sandeen.net>
Signed-off-by: NTim Shimmin <tes@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

Use the generic set, get and removexattr methods and supply the s_xattr
array with fine-grained handlers. All XFS/Linux highlevel attr handling is
rewritten from scratch and placed into fs/xfs/linux-2.6/xfs_xattr.c so
that it's separated from the generic low-level code.

SGI-PV: 982343

SGI-Modid: xfs-linux-melb:xfs-kern:31234a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NTim Shimmin <tes@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

The vfs_unlink/d_delete functionality in the Linux VFS make the
dentry negative if it is the only inode being referenced. Case-insensitive
mode doesn't work with negative dentries, so if using CI-mode, invalidate
the dentry on unlink/rmdir.

SGI-PV: 983102
SGI-Modid: xfs-linux-melb:xfs-kern:31308a
Signed-off-by: NBarry Naujok <bnaujok@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

Fixes a problem in the xfs_dir2_remove and xfs_dir2_replace paths which
intenally call directory format specific lookup funtions that assume
args->cmpresult is zeroed.

SGI-PV: 982606
SGI-Modid: xfs-linux-melb:xfs-kern:31268a
Signed-off-by: NBarry Naujok <bnaujok@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

SGI-PV: 981521
SGI-Modid: xfs-linux-melb:xfs-kern:31214a
Signed-off-by: NBarry Naujok <bnaujok@sgi.com>
Signed-off-by: NDavid Chinner <dgc@sgi.com>

SGI-PV: 981498
SGI-Modid: xfs-linux-melb:xfs-kern:31212a
Signed-off-by: NBarry Naujok <bnaujok@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

Implement ASCII case-insensitive support. It's primary purpose is for
supporting existing filesystems that already use this case-insensitive
mode migrated from IRIX. But, if you only need ASCII-only case-insensitive
support (ie. English only) and will never use another language, then this
mode is perfectly adequate.

ASCII-CI is implemented by generating hashes based on lower-case letters
and doing lower-case compares. It implements a new xfs_nameops vector for
doing the hashes and comparisons for all filename operations.

To create a filesystem with this CI mode, use: # mkfs.xfs -n version=ci
<device>

SGI-PV: 981516
SGI-Modid: xfs-linux-melb:xfs-kern:31209a
Signed-off-by: NBarry Naujok <bnaujok@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

This implements the code to store the actual filename found during a
lookup in the dentry cache and to avoid multiple entries in the dcache
pointing to the same inode.

To avoid polluting the dcache, we implement a new directory inode
operations for lookup. xfs_vn_ci_lookup() stores the correct case name in
the dcache.

The "actual name" is only allocated and returned for a case- insensitive
match and not an actual match.

Another unusual interaction with the dcache is not storing negative
dentries like other filesystems doing a d_add(dentry, NULL) when an ENOENT
is returned. During the VFS lookup, if a dentry returned has no inode,
dput is called and ENOENT is returned. By not doing a d_add, this actually
removes it completely from the dcache to be reused. create/rename have to
be modified to support unhashed dentries being passed in.

SGI-PV: 981521
SGI-Modid: xfs-linux-melb:xfs-kern:31208a
Signed-off-by: NBarry Naujok <bnaujok@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

This add a dcache entry to the dcache for lookup, but changing the name
that is associated with the entry rather than the one passed in to the
lookup routine.

First, it sees if the case-exact match already exists in the dcache and
uses it if one exists. Otherwise, it allocates a new node with the new
name and splices it into the dcache.

Original code from ntfs_lookup in fs/ntfs/namei.c by Anton Altaparmakov.
Signed-off-by: NBarry Naujok <bnaujok@sgi.com>
Signed-off-by: NAnton Altaparmakov <aia21@cantab.net>
Acked-by: NChristoph Hellwig <hch@infradead.org>

The end of the xfs_da_args structure has 4 unsigned char fields for
true/false information on directory and attr operations using the
xfs_da_args structure.

The following converts these 4 into a op_flags field that uses the first 4
bits for these fields and allows expansion for future operation
information (eg. case-insensitive lookup request).

SGI-PV: 981520
SGI-Modid: xfs-linux-melb:xfs-kern:31206a
Signed-off-by: NBarry Naujok <bnaujok@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

Adds two pieces of functionality for the basis of case-insensitive support
in XFS:

1. A comparison result enumerated type: xfs_dacmp. It represents an

exact match, case-insensitive match or no match at all. This patch

only implements different and exact results.

2. xfs_nameops vector for specifying how to perform the hash generation

of filenames and comparision methods. In this patch the hash vector

points to the existing xfs_da_hashname function and the comparison

method does a length compare, and if the same, does a memcmp and

return the xfs_dacmp result.

All filename functions that use the hash (create, lookup remove, rename,
etc) now use the xfs_nameops.hashname function and all directory lookup
functions also use the xfs_nameops.compname function.

The lookup functions also handle case-insensitive results even though the
default comparison function cannot return that. And important aspect of
the lookup functions is that an exact match always has precedence over a
case-insensitive. So while a case-insensitive match is found, we have to
keep looking just in case there is an exact match. In the meantime, the
info for the first case-insensitive match is retained if no exact match is
found.

SGI-PV: 981519
SGI-Modid: xfs-linux-melb:xfs-kern:31205a
Signed-off-by: NBarry Naujok <bnaujok@sgi.com>
Signed-off-by: NChristoph Hellwig <hch@infradead.org>

de-duplicate calls to xfs_attr_trace_enter

Every call to xfs_attr_trace_enter() shares the exact same 16 args in the
middle... just send in the context pointer and let the next level down
split it into the ktrace.

Compile tested only.

SGI-PV: 976035
SGI-Modid: xfs-linux-melb:xfs-kern:31200a
Signed-off-by: NEric Sandeen <sandeen@sandeen.net>
Signed-off-by: NNiv Sardi <xaiki@sgi.com>
Signed-off-by: NJosef 'Jeff' Sipek <jeffpc@josefsipek.net>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

SGI-PV: 981951
SGI-Modid: xfs-linux-melb:xfs-kern:31199a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NDavid Chinner <dgc@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

SGI-PV: 981951
SGI-Modid: xfs-linux-melb:xfs-kern:31198a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NDavid Chinner <dgc@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

xfs_binval aka xfs_flush_buftarg is the first thing done in
xfs_free_buftarg, so there is no need to have duplicated calls just before
xfs_free_buftarg in the mount failure path.

SGI-PV: 981951
SGI-Modid: xfs-linux-melb:xfs-kern:31197a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NDavid Chinner <dgc@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

xfs_mount_init is inlined into xfs_fs_fill_super and allocation switched
to kzalloc. Plug a leak of the mount structure for most early mount
failures. Move xfs_icsb_init_counters to as late as possible in the mount
path and make sure to undo it so that no stale hotplug cpu notifiers are
left around on mount failures.

SGI-PV: 981951
SGI-Modid: xfs-linux-melb:xfs-kern:31196a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NDavid Chinner <dgc@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

Switch xfs_args_allocate to kzalloc and handle failures.

SGI-PV: 981951
SGI-Modid: xfs-linux-melb:xfs-kern:31195a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NDavid Chinner <dgc@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

Split setting the block and sector size out of xfs_fs_fill_super into a
small helper to make xfs_fs_fill_super more readable.

SGI-PV: 981951
SGI-Modid: xfs-linux-melb:xfs-kern:31194a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NDavid Chinner <dgc@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

Currently closing the rt/log block device is done in the wrong spot, and
far too early. So revampt it:

- xfs_blkdev_put moved out of xfs_free_buftarg into the caller so that

it is done after tearing down the buftarg completely.

- call to xfs_unmountfs_close moved from xfs_mountfs into caller so

that it's done after tearing down the filesystem completely.

- xfs_unmountfs_close is renamed to xfs_close_devices and made static

in xfs_super.c

- opening of the block devices is split into a helper xfs_open_devices

that is symetric in use to xfs_close_devices

- xfs_unmountfs can now lose struct cred

- error handling around device opening sanitized in xfs_fs_fill_super

SGI-PV: 981951
SGI-Modid: xfs-linux-melb:xfs-kern:31193a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NDavid Chinner <dgc@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

Freeing of the superblock is already handled in the caller, and that is
more symmetric with the mount path, too.

SGI-PV: 981951
SGI-Modid: xfs-linux-melb:xfs-kern:31192a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NDavid Chinner <dgc@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

xfs_mount is already pretty linux-specific so merge it into
xfs_fs_fill_super to allow for a more structured mount code in the next
patches. xfs_start_flags and xfs_finish_flags also move to xfs_super.c.

SGI-PV: 981951
SGI-Modid: xfs-linux-melb:xfs-kern:31189a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NDavid Chinner <dgc@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

xfs_unmount is small and already pretty Linux specific, so merge it into
the callers. The real unmount path is simplified a little by doing a
WARN_ON on the xfs_unmount_flush retval directly instead of propagating
the error back to the caller, and the mout failure case in simplified
significantly by removing the forced shutdown case and all the dmapi
events that shouldn't be sent because the dmapi mount event hasn't been
sent by that time either.

SGI-PV: 981951
SGI-Modid: xfs-linux-melb:xfs-kern:31188a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NDavid Chinner <dgc@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

xfs_igrow_start just expands to xfs_zero_eof with two asserts that are
useless in the context of the only caller and some rather confusing
comments.

xfs_igrow_finish is just a few lines of code decorated again with useless
asserts and confusing comments.

Just kill those two and merge them into xfs_setattr.

SGI-PV: 981498
SGI-Modid: xfs-linux-melb:xfs-kern:31186a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NDavid Chinner <dgc@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>

xfs_mntupdate already is completely Linux specific due to the VFS flags
passed in, so it might aswell be merged into xfs_fs_remount.

SGI-PV: 981498
SGI-Modid: xfs-linux-melb:xfs-kern:31185a
Signed-off-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NDavid Chinner <dgc@sgi.com>
Signed-off-by: NLachlan McIlroy <lachlan@sgi.com>