That function is only called under nfsd_mutex: we know that because the
only caller is nfsd_svc, via

        nfsd_svc
          nfsd_startup
            nfs4_state_start
              nfsd4_client_tracking_init
                client_tracking_ops->init == nfsd4_load_reboot_recovery_data

The shared state accessed here includes:

        - user_recovery_dirname: used here, modified only by
          nfs4_reset_recoverydir, which can be verified to only be
          called under nfsd_mutex.
        - filesystem state, protected by i_mutex (handwaving slightly
	  here)
        - rec_file, reclaim_str_hashtbl, reclaim_str_hashtbl_size: other
          than here, used only from code called from nfsd or laundromat
          threads, both of which should be started only after this runs
          (see nfsd_svc) and stopped before this could run again (see
          nfsd_shutdown, called from nfsd_last_thread).
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

Passing net context looks as overkill.
Signed-off-by: NStanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

This hash holds nfs4_clients info, which are network namespace aware.
So let's make it allocated per network namespace.

Note: this hash is used only by legacy tracker. So let's allocate hash in
tracker init.
Signed-off-by: NStanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

And use it's net where possible.
Signed-off-by: NStanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

The current code holds on to this list until nfsd is shut down, but it's
never touched once the grace period ends. Release that memory back into
the wild when the grace period ends.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

Remove the cl_recdir field from the nfs4_client struct. Instead, just
compute it on the fly when and if it's needed, which is now only when
the legacy client tracking code is in effect.

The error handling in the legacy client tracker is also changed to
handle the case where md5 is unavailable. In that case, we'll warn
the admin with a KERN_ERR message and disable the client tracking.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

When nfsd starts, the legacy reboot recovery code creates a tracking
struct for each directory in the v4recoverydir. When the grace period
ends, it basically does a "readdir" on the directory again, and matches
each dentry in there to an existing client id to see if it should be
removed or not. If the matching client doesn't exist, or hasn't
reclaimed its state then it will remove that dentry.

This is pretty inefficient since it involves doing a lot of hash-bucket
searching. It also means that we have to keep relying on being able to
search for a nfs4_client by md5 hashed cl_recdir name.

Instead, add a pointer to the nfs4_client that indicates the association
between the nfs4_client_reclaim and nfs4_client. When a reclaim operation
comes in, we set the pointer to make that association. On gracedone, the
legacy client tracker will keep the recdir around iff:

1/ there is a reclaim record for the directory

...and...

2/ there's an association between the reclaim record and a client record
-- that is, a create or check operation was performed on the client that
matches that directory.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

Currently, it takes a client pointer, but later we're going to need to
search for these records without knowing whether a matching client even
exists.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

Let's shoot for removing the nfsdcld upcall in 3.10. Most likely,
no one is actually using it so I don't expect this warning to
fire often (except maybe on misconfigured systems).
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

The usermodehelper upcall program can then decide to use this info as
a (one-way) transition mechanism to the new scheme. When a "check"
upcall occurs and the client doesn't exist in the database, we can
look to see whether the directory exists. If it does, then we'd add
the client to the database, remove the legacy recdir, and return
success to the kernel to allow the recovery to proceed.

For gracedone, we simply pass the v4recovery "topdir" so that the
upcall can clean it out prior to returning to the kernel.

A module parm is also added to disable the legacy conversion if
the admin chooses.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

First, try to use the new usermodehelper upcall. It should succeed or
fail quickly, so there's little cost to doing so.

If it fails, and the legacy tracking dir exists, use that. If it
doesn't exist then fall back to using nfsdcld.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

Add a new client tracker upcall type that uses call_usermodehelper to
call out to a program. This seems to be the preferred method of
calling out to usermode these days for seldom-called upcalls. It's
simple and doesn't require a running daemon, so it should "just work"
as long as the binary is installed.

The client tracking exit operation is also changed to check for a
NULL pointer before running. The UMH upcall doesn't need to do anything
at module teardown time.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

If the credential save fails, then we'll leak our mnt_want_write_file
reference.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

When mnt_want_write() starts to handle freezing it will get a full lock
semantics requiring proper lock ordering. So push mnt_want_write() call
consistently outside of i_mutex.

CC: linux-nfs@vger.kernel.org
CC: "J. Bruce Fields" <bfields@fieldses.org>
Signed-off-by: NJan Kara <jack@suse.cz>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Reviewed-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

Fix printk format warnings -- both items are size_t,
so use %zu to print them.

fs/nfsd/nfs4recover.c:580:3: warning: format '%lu' expects type 'long unsigned int', but argument 3 has type 'size_t'
fs/nfsd/nfs4recover.c:580:3: warning: format '%lu' expects type 'long unsigned int', but argument 4 has type 'unsigned int'
Signed-off-by: NRandy Dunlap <rdunlap@xenotime.net>
Cc: "J. Bruce Fields" <bfields@fieldses.org>
Cc: linux-nfs@vger.kernel.org
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Otherwise, we get a warning or error similar to this when building with
CONFIG_NFSD_V4 disabled:

    ERROR: "nfsd4_cld_block" [fs/nfsd/nfsd.ko] undefined!

Fix this by wrapping the calls to rpc_pipefs_notifier_register and
..._unregister in another function and providing no-op replacements
when CONFIG_NFSD_V4 is disabled.
Reported-by: NPaul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

If we find that "cup" is NULL in this case, then we obviously don't
want to dereference it. What we really want to print in this case
is the xid that we copied off earlier.
Reported-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

This code isn't set up for containers, so don't allow it to be
used for anything but init_net.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

In the event that rpc_pipefs isn't mounted when nfsd starts, we
must register a notifier to handle creating the dentry once it
is mounted, and to remove the dentry on unmount.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

...and add a mechanism for switching between the "legacy" tracker and
the new one. The decision is made by looking to see whether the
v4recoverydir exists. If it does, then the legacy client tracker is
used.

If it's not, then the kernel will create a "cld" pipe in rpc_pipefs.
That pipe is used to talk to a daemon for handling the upcall.

Most of the data structures for the new client tracker are handled on a
per-namespace basis, so this upcall should be essentially ready for
containerization. For now however, nfsd just starts it by calling the
initialization and exit functions for init_net.

I'm making the assumption that at some point in the future we'll be able
to determine the net namespace from the nfs4_client. Until then, this
patch hardcodes init_net in those places. I've sprinkled some "FIXME"
comments around that code to attempt to make it clear where we'll need
to fix that up later.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

Abstract out the mechanism that we use to track clients into a set of
client name tracking functions.

This gives us a mechanism to plug in a new set of client tracking
functions without disturbing the callers. It also gives us a way to
decide on what tracking scheme to use at runtime.

For now, this just looks like pointless abstraction, but later we'll
add a new alternate scheme for tracking clients on stable storage.

Note too that this patch anticipates the eventual containerization
of this code by passing in struct net pointers in places. No attempt
is made to containerize the legacy client tracker however.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

We'll need a way to flag the nfs4_client as already being recorded on
stable storage so that we don't continually upcall. Currently, that's
recorded in the cl_firststate field of the client struct. Using an
entire u32 to store a flag is rather wasteful though.

The cl_cb_flags field is only using 2 bits right now, so repurpose that
to a generic flags field. Rename NFSD4_CLIENT_KILL to
NFSD4_CLIENT_CB_KILL to make it evident that it's part of the callback
flags. Add a mask that we can use for existing checks that look to see
whether any flags are set, so that the new flags don't interfere.

Convert all references to cl_firstate to the NFSD4_CLIENT_STABLE flag,
and add a new NFSD4_CLIENT_RECLAIM_COMPLETE flag.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

If the recovery directory doesn't exist, then behavior after a reboot
will be suboptimal.  But it's unnecessarily harsh to then prevent the
nfsv4 server from working at all.  Instead just print a warning
(already done in nfsd4_init_recdir()) and soldier on.
Tested-by: NLior <lior@tonian.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

new helper (wrapper around mnt_drop_write()) to be used in pair with
mnt_want_write_file().
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

it's both faster (in case when file has been opened for write) and cleaner.
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

In the NFSv4.1 case, this could cause a spurious "NFSD: failed to write
recovery record (err -17); please check that /var/lib/nfs/v4recovery
exists and is writable.
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>
Reported-by: NSteve Dickson <SteveD@redhat.com>

The server is returning nfserr_resource for both permanent errors and
for errors (like allocation failures) that might be resolved by retrying
later.  Save nfserr_resource for the former and use delay/jukebox for
the latter.

Cc: stable@kernel.org
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

Even if we fail to write a recovery record, we should still mark the
client as having acquired its first state.  Otherwise we leave 4.1
clients with indefinite ERR_GRACE returns.

However, an inability to write stable storage records may cause failures
of reboot recovery, and the problem should still be brought to the
server administrator's attention.

So, make sure the error is logged.

These errors shouldn't normally be triggered on a corectly functioning
server--this isn't a case where a misconfigured client could spam the
logs.
Signed-off-by: NBoaz Harrosh <bharrosh@panasas.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

Move around some of this code, simplify a bit.
Reviewed-by: NBoaz Harrosh <bharrosh@panasas.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

just rewind it to the beginning before vfs_readdir() and be
done with that...
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Instead of failing to find client entries which don't match the
minorversion, we should be finding them, then either erroring out or
expiring them as appropriate.

This also fixes a problem which would cause the 4.1 server to fail to
recognize clients after a second reboot.
Reported-by: NCasey Bodley <cbodley@citi.umich.edu>
Reviewed-by: NBenny Halevy <bhalevy@panasas.com>
Signed-off-by: NJ. Bruce Fields <bfields@redhat.com>

Now that the last user passing a NULL file pointer is gone we can remove
the redundant dentry argument and associated hacks inside vfs_fsynmc_range.

The next step will be removig the dentry argument from ->fsync, but given
the luck with the last round of method prototype changes I'd rather
defer this until after the main merge window.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

Instead of just looking up a path use do_filp_open to get us a file
structure for the nfs4 recovery directory.  This allows us to get
rid of the last non-standard vfs_fsync caller with a NULL file
pointer.

[AV: should be using fput(), not filp_close()]
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h

percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files.  percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.

percpu.h -> slab.h dependency is about to be removed.  Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability.  As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.

  http://userweb.kernel.org/~tj/misc/slabh-sweep.py

The script does the followings.

* Scan files for gfp and slab usages and update includes such that
  only the necessary includes are there.  ie. if only gfp is used,
  gfp.h, if slab is used, slab.h.

* When the script inserts a new include, it looks at the include
  blocks and try to put the new include such that its order conforms
  to its surrounding.  It's put in the include block which contains
  core kernel includes, in the same order that the rest are ordered -
  alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
  doesn't seem to be any matching order.

* If the script can't find a place to put a new include (mostly
  because the file doesn't have fitting include block), it prints out
  an error message indicating which .h file needs to be added to the
  file.

The conversion was done in the following steps.

1. The initial automatic conversion of all .c files updated slightly
   over 4000 files, deleting around 700 includes and adding ~480 gfp.h
   and ~3000 slab.h inclusions.  The script emitted errors for ~400
   files.

2. Each error was manually checked.  Some didn't need the inclusion,
   some needed manual addition while adding it to implementation .h or
   embedding .c file was more appropriate for others.  This step added
   inclusions to around 150 files.

3. The script was run again and the output was compared to the edits
   from #2 to make sure no file was left behind.

4. Several build tests were done and a couple of problems were fixed.
   e.g. lib/decompress_*.c used malloc/free() wrappers around slab
   APIs requiring slab.h to be added manually.

5. The script was run on all .h files but without automatically
   editing them as sprinkling gfp.h and slab.h inclusions around .h
   files could easily lead to inclusion dependency hell.  Most gfp.h
   inclusion directives were ignored as stuff from gfp.h was usually
   wildly available and often used in preprocessor macros.  Each
   slab.h inclusion directive was examined and added manually as
   necessary.

6. percpu.h was updated not to include slab.h.

7. Build test were done on the following configurations and failures
   were fixed.  CONFIG_GCOV_KERNEL was turned off for all tests (as my
   distributed build env didn't work with gcov compiles) and a few
   more options had to be turned off depending on archs to make things
   build (like ipr on powerpc/64 which failed due to missing writeq).

   * x86 and x86_64 UP and SMP allmodconfig and a custom test config.
   * powerpc and powerpc64 SMP allmodconfig
   * sparc and sparc64 SMP allmodconfig
   * ia64 SMP allmodconfig
   * s390 SMP allmodconfig
   * alpha SMP allmodconfig
   * um on x86_64 SMP allmodconfig

8. percpu.h modifications were reverted so that it could be applied as
   a separate patch and serve as bisection point.

Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: NTejun Heo <tj@kernel.org>
Guess-its-ok-by: NChristoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>

- Add commit_metadata export_operation to allow the underlying filesystem to
decide how to commit an inode most efficiently.

- Usage of nfsd_sync_dir and write_inode_now has been replaced with the
commit_metadata function that takes a svc_fh.

- The commit_metadata function calls the commit_metadata export_op if it's
there, or else falls back to sync_inode instead of fsync and write_inode_now
because only metadata need be synced here.

- nfsd4_sync_rec_dir now uses vfs_fsync so that commit_metadata can be static
Signed-off-by: NBen Myers <bpm@sgi.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

The new .h files have paths at the top that are now out of date.  While
we're here, just remove all of those from fs/nfsd; they never served any
purpose.
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>