Introduce a new data type, used by both the in-kernel NLM and NSM
implementations, that is used to manage the opaque "priv" argument
for the NSMPROC_MON and NLMPROC_SM_NOTIFY calls.

Construct the "priv" cookie when the nsm_handle is created.

The nsm_init_private() function may look a little strange, but it is
roughly equivalent to how the XDR encoder formed the "priv" argument.
It's going to go away soon.
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

The nsm_find() function sets up fresh nsm_handle entries.  This is
where we will store the "priv" cookie used to lookup nsm_handles during
reboot recovery.  The cookie will be constructed when nsm_find()
creates a new nsm_handle.

As much as possible, I would like to keep everything that handles a
"priv" cookie in fs/lockd/mon.c so that all the smarts are in one
source file.  That organization should make it pretty simple to see how
all this works.

To me, it makes more sense than the current arrangement to keep
nsm_find() with nsm_monitor() and nsm_unmonitor().

So, start reorganizing by moving nsm_find() into fs/lockd/mon.c.  The
nsm_release() function comes along too, since it shares the nsm_lock
global variable.
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Clean up: Move the RPC program and procedure numbers for NSM into the
one source file that needs them: fs/lockd/mon.c.

And, as with NLM, NFS, and rpcbind calls, use NSMPROC_FOO instead of
SM_FOO for NSM procedure numbers.

Finally, make a couple of comments more precise: what is referred to
here as SM_NOTIFY is really the NLM (lockd) NLMPROC_SM_NOTIFY downcall,
not NSMPROC_NOTIFY.
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Clean up: NSM's XDR data structures are used only in fs/lockd/mon.c,
so move them there.
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Clean up.

Make the nlm_host argument "const," and move the public declaration to
lockd.h.  Add a documenting comment.

Bruce observed that nsm_unmonitor()'s only caller doesn't care about
its return code, so make nsm_unmonitor() return void.
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

The nsm_handle's reference count is bumped in nlm_lookup_host().  It
should be decremented in nlm_destroy_host() to make it easier to see
the balance of these two operations.

Move the nsm_release() call to fs/lockd/host.c.

The h_nsmhandle pointer is set in nlm_lookup_host(), and never cleared.
The nlm_destroy_host() function is never called for the same nlm_host
twice, so h_nsmhandle won't ever be NULL when nsm_unmonitor() is
called.

All references to the nlm_host are gone before it is freed.  We can
skip making h_nsmhandle NULL just before the nlm_host is deallocated.

It's also likely we can remove the h_nsmhandle NULL check in
nlmsvc_is_client() as well, but we can do that later when rearchitect-
ing the nlm_host cache.
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Clean up.

Make the nlm_host argument "const," and move the public declaration to
lockd.h with other NSM public function (nsm_release, eg) and global
variable declarations.

Add a documenting comment.
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

The "mon_name" argument of the NSMPROC_MON and NSMPROC_UNMON upcalls
is a string that contains the hostname or IP address of the remote peer
to be notified when this host has rebooted.  The sm-notify command uses
this identifier to contact the peer when we reboot, so it must be
either a well-qualified DNS hostname or a presentation format IP
address string.

When the "nsm_use_hostnames" sysctl is set to zero, the kernel's NSM
provides a presentation format IP address in the "mon_name" argument.
Otherwise, the "caller_name" argument from NLM requests is used,
which is usually just the DNS hostname of the peer.

To support IPv6 addresses for the mon_name argument, we use the
nsm_handle's address eye-catcher, which already contains an appropriate
presentation format address string.  Using the eye-catcher string
obviates the need to use a large buffer on the stack to form the
presentation address string for the upcall.

This patch also addresses a subtle bug.

An NSMPROC_MON request and the subsequent NSMPROC_UNMON request for the
same peer are required to use the same value for the "mon_name"
argument.  Otherwise, rpc.statd's NSMPROC_UNMON processing cannot
locate the database entry for that peer and remove it.

If the setting of nsm_use_hostnames is changed between the time the
kernel sends an NSMPROC_MON request and the time it sends the
NSMPROC_UNMON request for the same peer, the "mon_name" argument for
these two requests may not be the same.  This is because the value of
"mon_name" is currently chosen at the moment the call is made based on
the setting of nsm_use_hostnames

To ensure both requests pass identical contents in the "mon_name"
argument, we now select which string to use for the argument in the
nsm_monitor() function.  A pointer to this string is saved in the
nsm_handle so it can be used for a subsequent NSMPROC_UNMON upcall.

NB: There are other potential problems, such as how nlm_host_rebooted()
might behave if nsm_use_hostnames were changed while hosts are still
being monitored.  This patch does not attempt to address those
problems.
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Clean up: I'm about to add another "char *" field to the nsm_handle
structure.  The sm_name field uses an older style of declaring a
"char *" field.  If I match that style for the new field, checkpatch.pl
will complain.

So, fix the sm_name field to use the new style.
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Scope ID support is needed since the kernel's NSM implementation is
about to use these displayed addresses as a mon_name in some cases.

When nsm_use_hostnames is zero, without scope ID support NSM will fail
to handle peers that contact us via a link-local address.  Link-local
addresses do not work without an interface ID, which is stored in the
sockaddr's sin6_scope_id field.
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

The h_name field in struct nlm_host is a just copy of
h_nsmhandle->sm_name.  Likewise, the contents of the h_addrbuf field
should be identical to the sm_addrbuf field.

The h_srcaddrbuf field is used only in one place for debugging.  We can
live without this until we get %pI formatting for printk().

Currently these buffers are 48 bytes, but we need to support scope IDs
in IPv6 presentation addresses, which means making the buffers even
larger.  Instead, let's find ways to eliminate them to save space.
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Clean up: I'm about to add another "char *" field to the nlm_host
structure.  The h_name field, for example, uses an older style of
declaring a "char *" field.  If I match that style for the new field,
checkpatch.pl will complain.

So, fix pointer fields to use the new style.
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

svc_check_conn_limits() attempts to prevent denial of service attacks
by having the service close old connections once it reaches a
threshold. This threshold is based on the number of threads in the
service:

	(serv->sv_nrthreads + 3) * 20

Once we reach this, we drop the oldest connections and a printk pops
to warn the admin that they should increase the number of threads.

Increasing the number of threads isn't an option however for services
like lockd. We don't want to eliminate this check entirely for such
services but we need some way to increase this limit.

This patch adds a sv_maxconn field to the svc_serv struct. When it's
set to 0, we use the current method to calculate the max number of
connections. RPC services can then set this on an as-needed basis.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Acked-by: NNeil Brown <neilb@suse.de>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Descriptions taken from mountd code (in nfs-utils/utils/mountd/cache.c).
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Implement barrier support for single device DM devices

This patch implements barrier support in DM for the common case of dm linear
just remapping a single underlying device. In this case we can safely
pass the barrier through because there can be no reordering between
devices.

 NB. Any DM device might cease to support barriers if it gets
     reconfigured so code must continue to allow for a possible
     -EOPNOTSUPP on every barrier bio submitted.  - agk
Signed-off-by: NAndi Kleen <ak@suse.de>
Signed-off-by: NMikulas Patocka <mpatocka@redhat.com>
Signed-off-by: NAlasdair G Kergon <agk@redhat.com>

This patch adds the following target interfaces for request-based dm.

  map_rq    : for mapping a request

  rq_end_io : for finishing a request

  busy      : for avoiding performance regression from bio-based dm.
              Target can tell dm core not to map requests now, and
              that may help requests in the block layer queue to be
              bigger by I/O merging.
              In bio-based dm, this behavior is done by device
              drivers managing the block layer queue.
              But in request-based dm, dm core has to do that
              since dm core manages the block layer queue.
Signed-off-by: NKiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: NJun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: NAlasdair G Kergon <agk@redhat.com>

Change dm_unregister_target to return void and use BUG() for error
reporting.

dm_unregister_target can only fail because of programming bug in the
target driver. It can't fail because of user's behavior or disk errors.

This patch changes unregister_target to return void and use BUG if
someone tries to unregister non-registered target or unregister target
that is in use.

This patch removes code duplication (testing of error codes in all dm
targets) and reports bugs in just one place, in dm_unregister_target. In
some target drivers, these return codes were ignored, which could lead
to a situation where bugs could be missed.
Signed-off-by: NMikulas Patocka <mpatocka@redhat.com>
Signed-off-by: NAlasdair G Kergon <agk@redhat.com>

An XFS workload showed up a bug in the lockless pagecache patch. Basically it
would go into an "infinite" loop, although it would sometimes be able to break
out of the loop! The reason is a missing compiler barrier in the "increment
reference count unless it was zero" case of the lockless pagecache protocol in
the gang lookup functions.

This would cause the compiler to use a cached value of struct page pointer to
retry the operation with, rather than reload it. So the page might have been
removed from pagecache and freed (refcount==0) but the lookup would not correctly
notice the page is no longer in pagecache, and keep attempting to increment the
refcount and failing, until the page gets reallocated for something else. This
isn't a data corruption because the condition will be detected if the page has
been reallocated. However it can result in a lockup.

Linus points out that ACCESS_ONCE is also required in that pointer load, even
if it's absence is not causing a bug on our particular build. The most general
way to solve this is just to put an rcu_dereference in radix_tree_deref_slot.

Assembly of find_get_pages,
before:
.L220:
        movq    (%rbx), %rax    #* ivtmp.1162, tmp82
        movq    (%rax), %rdi    #, prephitmp.1149
.L218:
        testb   $1, %dil        #, prephitmp.1149
        jne     .L217   #,
        testq   %rdi, %rdi      # prephitmp.1149
        je      .L203   #,
        cmpq    $-1, %rdi       #, prephitmp.1149
        je      .L217   #,
        movl    8(%rdi), %esi   # <variable>._count.counter, c
        testl   %esi, %esi      # c
        je      .L218   #,

after:
.L212:
        movq    (%rbx), %rax    #* ivtmp.1109, tmp81
        movq    (%rax), %rdi    #, ret
        testb   $1, %dil        #, ret
        jne     .L211   #,
        testq   %rdi, %rdi      # ret
        je      .L197   #,
        cmpq    $-1, %rdi       #, ret
        je      .L211   #,
        movl    8(%rdi), %esi   # <variable>._count.counter, c
        testl   %esi, %esi      # c
        je      .L212   #,

(notice the obvious infinite loop in the first example, if page->count remains 0)
Signed-off-by: NNick Piggin <npiggin@suse.de>
Reviewed-by: NPaul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The problems lie in the types used for some inotify interfaces, both at the kernel level and at the glibc level. This mail addresses the kernel problem. I will follow up with some suggestions for glibc changes.

For the sys_inotify_rm_watch() interface, the type of the 'wd' argument is
currently 'u32', it should be '__s32' .  That is Robert's suggestion, and
is consistent with the other declarations of watch descriptors in the
kernel source, in particular, the inotify_event structure in
include/linux/inotify.h:

struct inotify_event {
        __s32           wd;             /* watch descriptor */
        __u32           mask;           /* watch mask */
        __u32           cookie;         /* cookie to synchronize two events */
        __u32           len;            /* length (including nulls) of name */
        char            name[0];        /* stub for possible name */
};

The patch makes the changes needed for inotify_rm_watch().
Signed-off-by: NMichael Kerrisk <mtk.manpages@googlemail.com>
Cc: Robert Love <rlove@google.com>
Cc: Vegard Nossum <vegard.nossum@gmail.com>
Cc: Ulrich Drepper <drepper@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Fsync currently has a fdatawrite/fdatawait pair around the method call,
and a mutex_lock/unlock of the inode mutex.  All callers of fsync have
to duplicate this, but we have a few and most of them don't quite get
it right.  This patch adds a new vfs_fsync that takes care of this.
It's a little more complicated as usual as ->fsync might get a NULL file
pointer and just a dentry from nfsd, but otherwise gets afile and we
want to take the mapping and file operations from it when it is there.

Notes on the fsync callers:

 - ecryptfs wasn't calling filemap_fdatawrite / filemap_fdatawait on the
   	lower file
 - coda wasn't calling filemap_fdatawrite / filemap_fdatawait on the host
	file, and returning 0 when ->fsync was missing
 - shm wasn't calling either filemap_fdatawrite / filemap_fdatawait nor
   taking i_mutex.  Now given that shared memory doesn't have disk
   backing not doing anything in fsync seems fine and I left it out of
   the vfs_fsync conversion for now, but in that case we might just
   not pass it through to the lower file at all but just call the no-op
   simple_sync_file directly.

[and now actually export vfs_fsync]
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Filesystems often to do compute intensive operation on some
metadata.  If this operation is repeated many times, it can be very
expensive.  It would be much nicer if the operation could be performed
once before a buffer goes to disk.

This adds triggers to jbd2 buffer heads.  Just before writing a metadata
buffer to the journal, jbd2 will optionally call a commit trigger associated
with the buffer.  If the journal is aborted, an abort trigger will be
called on any dirty buffers as they are dropped from pending
transactions.

ocfs2 will use this feature.

Initially I tried to come up with a more generic trigger that could be
used for non-buffer-related events like transaction completion.  It
doesn't tie nicely, because the information a buffer trigger needs
(specific to a journal_head) isn't the same as what a transaction
trigger needs (specific to a tranaction_t or perhaps journal_t).  So I
implemented a buffer set, with the understanding that
journal/transaction wide triggers should be implemented separately.

There is only one trigger set allowed per buffer.  I can't think of any
reason to attach more than one set.  Contrast this with a journal or
transaction in which multiple places may want to watch the entire
transaction separately.

The trigger sets are considered static allocation from the jbd2
perspective.  ocfs2 will just have one trigger set per block type,
setting the same set on every bh of the same type.
Signed-off-by: NJoel Becker <joel.becker@oracle.com>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: NMark Fasheh <mfasheh@suse.com>

These are default functions for creating and destroying quota structures
and they should be used from filesystems.
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NMark Fasheh <mfasheh@suse.com>

Unexport header files dqblk_v[12].h since except for quota format ID they
don't contain information userspace should be interested in. Move ID
definitions to quota.h.
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NMark Fasheh <mfasheh@suse.com>

Add this so that file systems using JBD2 can safely allocate unused b_state
bits.

In this case, we add it so that Ocfs2 can define a single bit for tracking
the validation state of a buffer.
Acked-by: N"Theodore Ts'o" <tytso@mit.edu>
Signed-off-by: NMark Fasheh <mfasheh@suse.com>

OCFS2 needs to scan all active dquots once in a while and sync quota
information among cluster nodes. Provide a helper function for it so
that it does not have to reimplement internally a list which VFS
already has. Moreover this function is probably going to be useful
for other clustered filesystems if they decide to use VFS quotas.
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NMark Fasheh <mfasheh@suse.com>

OCFS2 needs to peek whether quota structure is already in memory so
that it can avoid expensive cluster locking in that case. Similarly
when freeing dquots, it checks whether it is the last quota structure
user or not. Finally, it needs to get reference to dquot structure for
specified id and quota type when recovering quota file after crash.
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NMark Fasheh <mfasheh@suse.com>

Increase reported version number of quota support since quota core has changed
significantly. Also remove __DQUOT_NUM_VERSION__ since nobody uses it.
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NMark Fasheh <mfasheh@suse.com>

Quota in a clustered environment needs to synchronize quota information
among cluster nodes. This means we have to occasionally update some
information in dquot from disk / network. On the other hand we have to
be careful not to overwrite changes administrator did via SETQUOTA.
So indicate in dquot->dq_flags which entries have been set by SETQUOTA
and quota format can clear these flags when it properly propagated
the changes.
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NMark Fasheh <mfasheh@suse.com>

For clustered filesystems, it can happen that space / inode usage goes
negative temporarily (because some node is allocating another node
is freeing and they are not completely in sync). So let quota code
allow this and change qsize_t so a signed type so that we don't
underflow the variables.
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NMark Fasheh <mfasheh@suse.com>

Coming quota support for OCFS2 is going to need quite a bit
of additional per-sb quota information. Moreover having fs.h
include all the types needed for this structure would be a
pain in the a**. So remove the union from mem_dqinfo and add
a private pointer for filesystem's use.
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NMark Fasheh <mfasheh@suse.com>

There is going to be a new version of quota format having 64-bit
quota limits and a new quota format for OCFS2. They are both
going to use the same tree structure as VFSv0 quota format. So
split out tree handling into a separate file and make size of
leaf blocks, amount of space usable in each block (needed for
checksumming) and structures contained in them configurable
so that the code can be shared.
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NMark Fasheh <mfasheh@suse.com>

Since these include files are used only by implementation of quota formats,
there's no need to have them in include/linux/.
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NMark Fasheh <mfasheh@suse.com>

If filesystem can handle quota files as system files hidden from users, we can
skip a lot of cache invalidation, syncing, inode flags setting etc. when
turning quotas on, off and quota_sync. Allow filesystem to indicate that it is
hiding quota files from users by DQUOT_QUOTA_SYS_FILE flag.
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NMark Fasheh <mfasheh@suse.com>

Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NMark Fasheh <mfasheh@suse.com>

Split DQUOT_USR_ENABLED (and DQUOT_GRP_ENABLED) into DQUOT_USR_USAGE_ENABLED
and DQUOT_USR_LIMITS_ENABLED. This way we are able to separately enable /
disable whether we should:
1) ignore quotas completely
2) just keep uptodate information about usage
3) actually enforce quota limits

This is going to be useful when quota is treated as filesystem metadata - we
then want to keep quota information uptodate all the time and just enable /
disable limits enforcement.
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NMark Fasheh <mfasheh@suse.com>

Upto now, DQUOT_USR_SUSPENDED behaved like a state - i.e., either quota
was enabled or suspended or none. Now allowed states are 0, ENABLED,
ENABLED | SUSPENDED. This will be useful later when we implement separate
enabling of quota usage tracking and limits enforcement because we need to
keep track of a state which has been suspended.
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NMark Fasheh <mfasheh@suse.com>

So far quota was fine with quota block limits and inode limits/numbers in
a 32-bit type. Now with rapid increase in storage sizes there are coming
requests to be able to handle quota limits above 4TB / more that 2^32 inodes.
So bump up sizes of types in mem_dqblk structure to 64-bits to be able to
handle this. Also update inode allocation / checking functions to use qsize_t
and make global structure keep quota limits in bytes so that things are
consistent.
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NMark Fasheh <mfasheh@suse.com>

Some filesystems would like to keep private information together with each
dquot. Add callbacks alloc_dquot and destroy_dquot allowing filesystem to
allocate larger dquots from their private slab in a similar fashion we
currently allocate inodes.
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NMark Fasheh <mfasheh@suse.com>

This patch implements the FIEMAP ioctl for GFS2. We can use the generic
code (aside from a lock order issue, solved as per Ted Tso's suggestion)
for which I've introduced a new variant of the generic function. We also
have one exception to deal with, namely stuffed files, so we do that
"by hand", setting all the required flags.

This has been tested with a modified (I could only find an old version) of
Eric's test program, and appears to work correctly.

This patch does not currently support FIEMAP of xattrs, but the plan is to add
that feature at some future point.
Signed-off-by: NSteven Whitehouse <swhiteho@redhat.com>
Cc: Theodore Tso <tytso@mit.edu>
Cc: Eric Sandeen <sandeen@redhat.com>

This patch allows GRO to merge page frags (skb_shinfo(skb)->frags)
in one skb, rather than using the less efficient frag_list.

It also adds a new interface, napi_gro_frags to allow drivers
to inject page frags directly into the stack without allocating
an skb.  This is intended to be the GRO equivalent for LRO's
lro_receive_frags interface.

The existing GSO interface can already handle page frags with
or without an appended frag_list so nothing needs to be changed
there.

The merging itself is rather simple.  We store any new frag entries
after the last existing entry, without checking whether the first
new entry can be merged with the last existing entry.  Making this
check would actually be easy but since no existing driver can
produce contiguous frags anyway it would just be mental masturbation.

If the total number of entries would exceed the capacity of a
single skb, we simply resort to using frag_list as we do now.
Signed-off-by: NHerbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: NDavid S. Miller <davem@davemloft.net>