The names of the callback functions used for virtual extended
attributes are based only on the last component of the attribute
name.  Because of the way these are defined, this precludes allowing
a single (lowest) attribute name for different callbacks, dependent
on the type of file being operated on.  (For example, it might be
nice to support both "ceph.dir.layout" and "ceph.file.layout".)

Just change the callback names to avoid this problem.
Signed-off-by: NAlex Elder <elder@dreamhost.com>
Signed-off-by: NSage Weil <sage@newdream.net>

A struct ceph_vxattr_cb does not represent a callback at all, but
rather a virtual extended attribute itself.  Drop the "_cb" suffix
from its name to reflect that.
Signed-off-by: NAlex Elder <elder@dreamhost.com>
Signed-off-by: NSage Weil <sage@newdream.net>

Entries in the ceph virtual extended attribute tables all follow a
distinct pattern in their definition.  Enforce this pattern through
the use of a macro.

Also, a null name field signals the end of the table, so make that
be the first field in the ceph_vxattr_cb structure.
Signed-off-by: NAlex Elder <elder@dreamhost.com>
Signed-off-by: NSage Weil <sage@newdream.net>

Use symbolic constants to define the top-level prefix for "ceph."
extended attribute names.
Signed-off-by: NAlex Elder <elder@dreamhost.com>
Signed-off-by: NSage Weil <sage@newdream.net>

All callers of ceph_match_vxattr() determine what to pass as the
first argument by calling ceph_inode_vxattrs(inode).  Just do that
inside ceph_match_vxattr() itself, changing it to take an inode
rather than the vxattr pointer as its first argument.

Also ensure the function works correctly for an empty table (i.e.,
containing only a terminating null entry).
Signed-off-by: NAlex Elder <elder@dreamhost.com>
Signed-off-by: NSage Weil <sage@newdream.net>

For some reason, ceph_setxattr() allocates an extra byte in which a
'\0' is stored past the end of an extended attribute value.  This is
not needed, and is potentially misleading, so get rid of it.
Signed-off-by: NAlex Elder <elder@dreamhost.com>
Signed-off-by: NSage Weil <sage@newdream.net>

The overflow check for a + n * b should be (n > (ULONG_MAX - a) / b),
rather than (n > ULONG_MAX / b - a).
Signed-off-by: NXi Wang <xi.wang@gmail.com>
Signed-off-by: NSage Weil <sage@newdream.net>

Return -EINVAL rather than panic if iinfo->symlink_len and inode->i_size
do not match.

Also use kstrndup rather than kmalloc/memcpy.
Signed-off-by: NXi Wang <xi.wang@gmail.com>
Reviewed-by: NAlex Elder <elder@dreamhost.com>

The root directory of the Ceph mount has inode number 1, so falling back
to 1 always creates a collision. 2 is unused on my test systems and seems
less likely to collide.
Signed-off-by: NAmon Ott <ao@m-privacy.de>
Signed-off-by: NSage Weil <sage@newdream.net>

Avoid the need to check for a special zero s_cap_ttl value by just
using (jiffies - 1) as the value assigned to indicate "sometime in
the past."
Signed-off-by: NAlex Elder <elder@dreamhost.com>
Reviewed-by: NSage Weil <sage@newdream.net>

Lockdep was reporting a possible circular lock dependency in
dentry_lease_is_valid().  That function needs to sample the
session's s_cap_gen and and s_cap_ttl fields coherently, but needs
to do so while holding a dentry lock.  The s_cap_lock field was
being used to protect the two fields, but that can't be taken while
holding a lock on a dentry within the session.

In most cases, the s_cap_gen and s_cap_ttl fields only get operated
on separately.  But in three cases they need to be updated together.
Implement a new lock to protect the spots updating both fields
atomically is required.
Signed-off-by: NAlex Elder <elder@dreamhost.com>
Reviewed-by: NSage Weil <sage@newdream.net>

"len" is read from network and thus needs validation.  Otherwise, given
a bogus "len" value, p+len could be an out-of-bounds pointer, which is
used in further parsing.
Signed-off-by: NXi Wang <xi.wang@gmail.com>
Signed-off-by: NSage Weil <sage@newdream.net>

The virtual extended attribute named "ceph.layout" is meaningful
only for regular files.  Change its name to be "ceph.file.layout" to
more directly reflect that in the ceph xattr namespace.  Preserve
the old "ceph.layout" name for the time being (until we decide it's
safe to get rid of it entirely).

Add a missing initializer for "readonly" in the terminating entry.
Signed-off-by: NAlex Elder <elder@dreamhost.com>
Reviewed-by: NSage Weil <sage@newdream.net>

In __ceph_build_xattrs_blob(), if a ceph inode's extended attributes
are marked dirty, all attributes recorded in its rb_tree index are
formatted into a "blob" buffer.  The target buffer is recorded in
ceph_inode->i_xattrs.prealloc_blob, and it is expected to exist and
be of sufficient size to hold the attributes.

The extended attributes are marked dirty in two cases: when a new
attribute is added to the inode; or when one is removed.  In the
former case work is done to ensure the prealloc_blob buffer is
properly set up, but in the latter it is not.

Change the logic in ceph_removexattr() so it matches what is
done in ceph_setxattr().  Note that this is done in a way that
keeps the two blocks of code nearly identical, in anticipation
of a subsequent patch that encapsulates some of this logic into
one or more helper routines.
Signed-off-by: NAlex Elder <elder@dreamhost.com>
Signed-off-by: NSage Weil <sage@newdream.net>

Enable/disable use of the dentry dir 'complete' flag via a mount option.
This lets the admin control whether ceph uses the dcache to satisfy
negative lookups or readdir when it has the entire directory contents in
its cache.

This is purely a performance optimization; correctness is guaranteed
whether it is enabled or not.
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NSage Weil <sage@newdream.net>

When open_root_dentry() gets a dentry via d_obtain_alias() it does
not get initialized.  If the dentry obtained came from the cache,
this is OK.  But if not, the result is an improperly initialized
dentry.

To fix this, call ceph_init_dentry() regardless of which path
produced the dentry.  That function returns immediately for a dentry
that is already initialized, it is safe to use either way.

(Credit to Sage, who suggested this fix.)
Signed-off-by: NAlex Elder <aelder@sgi.com>

ceph_mdsc_put_request() can call iput(), which can sleep.  Don't do that.

Fixes: #1812
Signed-off-by: NSage Weil <sage@newdream.net>

Nothing we do here sleeps, so just do it under d_lock and avoid the dget/
dput entirely.
Reported-by: NAl Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: NSage Weil <sage@newdream.net>

moved dereference after BUG_ON
Signed-off-by: NYehuda Sadeh <yehuda@hq.newdream.net>

We now set d_fsdata unconditionally on all dentries prior to setting up
the d_ops, so all of these checks are unnecessary.
Signed-off-by: NSage Weil <sage@newdream.net>

... and ceph_init_dentry(NULL) will oops
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

vfs_create() ignores everything outside of 16bit subset of its
mode argument; switching it to umode_t is obviously equivalent
and it's the only caller of the method
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

vfs_mkdir() gets int, but immediately drops everything that might not
fit into umode_t and that's the only caller of ->mkdir()...
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Seeing that just about every destructor got that INIT_LIST_HEAD() copied into
it, there is no point whatsoever keeping this INIT_LIST_HEAD in inode_init_once();
the cost of taking it into inode_init_always() will be negligible for pipes
and sockets and negative for everything else.  Not to mention the removal of
boilerplate code from ->destroy_inode() instances...
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Ceph attempts to use the dcache to satisfy negative lookups and readdir
when the entire directory contents are in cache.  Disable this behavior
until lingering bugs in this code are shaken out; we'll re-enable these
hooks once things are fully stable.
Signed-off-by: NSage Weil <sage@newdream.net>

one of the paths was missing spin_unlock
Signed-off-by: NYehuda Sadeh <yehuda@hq.newdream.net>

Commit 06222e49 got the if wrong so that
it always evaluates as true.  This is semantically harmless, but makes
SEEK_CUR and SEEK_SET needlessly query the server.

Rewrite the if to explicitly enumerate the cases we DO need a valid i_size
to make this code less fragile.
Reported-by: NRoel Kluin <roel.kluin@gmail.com>
Signed-off-by: NSage Weil <sage@newdream.net>

We have been using i_lock to protect all kinds of data structures in the
ceph_inode_info struct, including lists of inodes that we need to iterate
over while avoiding races with inode destruction.  That requires grabbing
a reference to the inode with the list lock protected, but igrab() now
takes i_lock to check the inode flags.

Changing the list lock ordering would be a painful process.

However, using a ceph-specific i_ceph_lock in the ceph inode instead of
i_lock is a simple mechanical change and avoids the ordering constraints
imposed by igrab().
Reported-by: NAmon Ott <a.ott@m-privacy.de>
Signed-off-by: NSage Weil <sage@newdream.net>

Fix typo.
Reported-by: Nmowang da <whooya.xxl@gmail.com>
Signed-off-by: NSage Weil <sage@newdream.net>

Set up d_fsdata on the root dentry.  This fixes a NULL pointer dereference
in ceph_d_prune on umount.  It also means we can eventually strip out all
of the conditional checks on d_fsdata because it is now set unconditionally
(prior to setting up the d_ops).

Fix the ceph_d_prune debug print while we're here.
Signed-off-by: NSage Weil <sage@newdream.net>

If we queue a work item that calls iput(), make sure we ihold() before
attempting to queue work. Otherwise our queued work might miraculously run
before we notice the queue_work() succeeded and call ihold(), allowing the
inode to be destroyed.

That is, instead of

	if (queue_work(...))
		ihold();

we need to do

	ihold();
	if (!queue_work(...))
		iput();
Reported-by: NAmon Ott <a.ott@m-privacy.de>
Signed-off-by: NSage Weil <sage@newdream.net>

Quiet the sparse noise:

warning: symbol 'create_fs_client' was not declared. Should it be static?
warning: symbol 'destroy_fs_client' was not declared. Should it be static?
Signed-off-by: NH Hartley Sweeten <hsweeten@visionengravers.com>
Cc: Sage Weil <sage@newdream.net>
ceph-devel@vger.kernel.org
Signed-off-by: NSage Weil <sage@newdream.net>

Quiet the following sparse noise:

warning: symbol 'get_nonsnap_parent' was not declared. Should it be static?
warning: symbol 'done_closing_sessions' was not declared. Should it be static?

Local functions don't need external visability. Make them static.
Signed-off-by: NH Hartley Sweeten <hsweeten@visionengravers.com>
Cc: Sage Weil <sage@newdream.net>
Signed-off-by: NSage Weil <sage@newdream.net>

We used to use a flag on the directory inode to track whether the dcache
contents for a directory were a complete cached copy.  Switch to a dentry
flag CEPH_D_COMPLETE that is safely updated by ->d_prune().
Signed-off-by: NSage Weil <sage@newdream.net>

When the VFS prunes a dentry from the cache, clear the D_COMPLETE flag
on the parent dentry.  Do this for the live and snapshotted namespaces. Do
not bother for the .snap dir contents, since we do not cache that.
Signed-off-by: NSage Weil <sage@newdream.net>

Replace remaining direct i_nlink updates with a new set_nlink()
updater function.
Signed-off-by: NMiklos Szeredi <mszeredi@suse.cz>
Tested-by: NToshiyuki Okajima <toshi.okajima@jp.fujitsu.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>