testing with newpynfs revealed this warning:
Jul  3 07:32:50 buml kernel: writeable file with no mnt_want_write()
Jul  3 07:32:50 buml kernel: ------------[ cut here ]------------
Jul  3 07:32:50 buml kernel: WARNING: at /usr0/export/dev/bhalevy/git/linux-pnfs-bh-nfs41/include/linux/fs.h:855 drop_file_write_access+0x6b/0x7e()
Jul  3 07:32:50 buml kernel: Modules linked in: nfsd auth_rpcgss exportfs nfs lockd nfs_acl sunrpc
Jul  3 07:32:50 buml kernel: Call Trace:
Jul  3 07:32:50 buml kernel: 6eaadc88:  [<6002f471>] warn_on_slowpath+0x54/0x8e
Jul  3 07:32:50 buml kernel: 6eaadcc8:  [<601b790d>] printk+0xa0/0x793
Jul  3 07:32:50 buml kernel: 6eaadd38:  [<601b6205>] __mutex_lock_slowpath+0x1db/0x1ea
Jul  3 07:32:50 buml kernel: 6eaadd68:  [<7107d4d5>] nfs4_preprocess_seqid_op+0x2a6/0x31c [nfsd]
Jul  3 07:32:50 buml kernel: 6eaadda8:  [<60078dc9>] drop_file_write_access+0x6b/0x7e
Jul  3 07:32:50 buml kernel: 6eaaddc8:  [<710804e4>] nfsd4_open_downgrade+0x114/0x1de [nfsd]
Jul  3 07:32:50 buml kernel: 6eaade08:  [<71076215>] nfsd4_proc_compound+0x1ba/0x2dc [nfsd]
Jul  3 07:32:50 buml kernel: 6eaade48:  [<71068221>] nfsd_dispatch+0xe5/0x1c2 [nfsd]
Jul  3 07:32:50 buml kernel: 6eaade88:  [<71312f81>] svc_process+0x3fd/0x714 [sunrpc]
Jul  3 07:32:50 buml kernel: 6eaadea8:  [<60039a81>] kernel_sigprocmask+0xf3/0x100
Jul  3 07:32:50 buml kernel: 6eaadee8:  [<7106874b>] nfsd+0x182/0x29b [nfsd]
Jul  3 07:32:50 buml kernel: 6eaadf48:  [<60021cc9>] run_kernel_thread+0x41/0x4a
Jul  3 07:32:50 buml kernel: 6eaadf58:  [<710685c9>] nfsd+0x0/0x29b [nfsd]
Jul  3 07:32:50 buml kernel: 6eaadf98:  [<60021cb0>] run_kernel_thread+0x28/0x4a
Jul  3 07:32:50 buml kernel: 6eaadfc8:  [<60013829>] new_thread_handler+0x72/0x9c
Jul  3 07:32:50 buml kernel:
Jul  3 07:32:50 buml kernel: ---[ end trace 2426dd7cb2fba3bf ]---

Bruce Fields suggested this (Thanks!):
maybe we need to be doing a mnt_want_write on open_upgrade and mnt_put_write on downgrade?

This patch adds a call to mnt_want_write and file_take_write (which is
doing the actual work).

The counter-calls mnt_drop_write a file_release_write are now being properly
called by drop_file_write_access in the exact path printed by the warning
above.
Signed-off-by: NBenny Halevy <bhalevy@panasas.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

It's not immediately obvious from the code why we're doing this.
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>
Cc: Benny Halevy <bhalevy@panasas.com>

In preparation for minorversion 1

All encoders now return an nfserr status (typically their
nfserr argument).  Unsupported ops go through nfsd4_encode_operation
too, so use nfsd4_encode_noop to encode nothing for their reply body.
Signed-off-by: NBenny Halevy <bhalevy@panasas.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Signed-off-by: NBenny Halevy <bhalevy@panasas.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Have separate vectors of operation decoders for each minorversion.
Obsolete ops in newer minorversions have default implementation returning
nfserr_opnotsupp.
Signed-off-by: NBenny Halevy <bhalevy@panasas.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

nfserr_opnotsupp should be returned for unsupported nfs4 ops
rather than nfserr_op_illegal.
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

In preparation for minorversion 1
Signed-off-by: NBenny Halevy <bhalevy@panasas.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Check minorversion once before decoding any operation and reject with
nfserr_minor_vers_mismatch if != 0 (this still happens in nfsd4_proc_compound).
In this case return a zero length resultdata array as required by RFC3530.

minorversion 1 processing will have its own vector of decoders.
Signed-off-by: NBenny Halevy <bhalevy@panasas.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Multiple mnt_want_write() calls in the switch statement looks really
ugly.
Signed-off-by: NMiklos Szeredi <mszeredi@suse.cz>
Acked-by: NChristoph Hellwig <hch@infradead.org>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

knfsd currently uses 2 signal masks when processing requests. A "loose"
mask (SHUTDOWN_SIGS) that it uses when receiving network requests, and
then a more "strict" mask (ALLOWED_SIGS, which is just SIGKILL) that it
allows when doing the actual operation on the local storage.

This is apparently unnecessarily complicated. The underlying filesystem
should be able to sanely handle a signal in the middle of an operation.
This patch removes the signal mask handling from knfsd altogether. When
knfsd is started as a kthread, all signals are ignored. It then allows
all of the signals in SHUTDOWN_SIGS. There's no need to set the mask
as well.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Thanks to Frank Van Maarseveen for the original problem report: "A
privileged process on an NFS client which drops privileges after using
them to change the current working directory, will experience incorrect
EACCES after an NFS server reboot. This problem can also occur after
memory pressure on the server, particularly when the client side is
quiet for some time."

This occurs because the filehandle points to a directory whose parents
are no longer in the dentry cache, and we're attempting to reconnect the
directory to its parents without adequate permissions to perform lookups
in the parent directories.

We can therefore fix the problem by acquiring the necessary capabilities
before attempting the reconnection.  We do this only in the
no_subtree_check case, since the documented behavior of the
subtree_check export option requires the server to check that the user
has lookup permissions on all parents.

The subtree_check case still has a problem, since reconnect_path()
unnecessarily requires both read and lookup permissions on all parent
directories.  However, a fix in that case would be more delicate, and
use of subtree_check is already discouraged for other reasons.
Signed-off-by: NNeil Brown <neilb@suse.de>
Cc: Frank van Maarseveen <frankvm@frankvm.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Rename nfsd_permission() specific MAY_* flags to NFSD_MAY_* to make it
clear, that these are not used outside nfsd, and to avoid name and
number space conflicts with the VFS.

[comment from hch: rename MAY_READ, MAY_WRITE and MAY_EXEC as well]
Signed-off-by: NMiklos Szeredi <mszeredi@suse.cz>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

OCFS2 can return -ERESTARTSYS from write requests (and possibly
elsewhere) if there is a signal pending.

If nfsd is shutdown (by sending a signal to each thread) while there
is still an IO load from the client, each thread could handle one last
request with a signal pending.  This can result in -ERESTARTSYS
which is not understood by nfserrno() and so is reflected back to
the client as nfserr_io aka -EIO.  This is wrong.

Instead, interpret ERESTARTSYS to mean "try again later" by returning
nfserr_jukebox.  The client will resend and - if the server is
restarted - the write will (hopefully) be successful and everyone will
be happy.

 The symptom that I narrowed down to this was:
    copy a large file via NFS to an OCFS2 filesystem, and restart
    the nfs server during the copy.
    The 'cp' might get an -EIO, and the file will be corrupted -
    presumably holes in the middle where writes appeared to fail.
Signed-off-by: NNeil Brown <neilb@suse.de>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

We need the nfsd_mutex before accessing nfsd_serv->sv_nrthreads or we
can't even guarantee nfsd_serv will still be there.
Signed-off-by: NNeil Brown <neilb@suse.de>
Acked-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Since we no longer make any distinction between shutdown signals with
nfsd, then it becomes easier to just standardize on a particular signal
to use to bring it down (SIGINT, in this case).
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

This patch is rather large, but I couldn't figure out a way to break it
up that would remain bisectable. It does several things:

- change svc_thread_fn typedef to better match what kthread_create expects
- change svc_pool_map_set_cpumask to be more kthread friendly. Make it
  take a task arg and and get rid of the "oldmask"
- have svc_set_num_threads call kthread_create directly
- eliminate __svc_create_thread
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

The special handling for SIGHUP in knfsd is a holdover from much
earlier versions of Linux where reloading the export table was
more expensive. That facility is not really needed anymore and
to my knowledge, is seldom-used.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Several of the nfsd filesystem interfaces allow changes to parameters
that don't have any effect on a running nfsd service. They are only ever
checked when nfsd is started. This patch fixes it so that changes to
those procfiles return -EBUSY if nfsd is already running to make it
clear that changes on the fly don't work.

The patch should also close some relatively harmless races between
changing the info in those interfaces and starting nfsd, since these
variables are being moved under the protection of the nfsd_mutex.

Finally, the nfsv4recoverydir file always returns -EINVAL if read. This
patch fixes it to return the recoverydir path as expected.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

This removes the BKL from the RPC service creation codepath. The BKL
really isn't adequate for this job since some of this info needs
protection across sleeps.

Also, add some comments to try and clarify how the locking should work
and to make it clear that the BKL isn't necessary as long as there is
adequate locking between tasks when touching the svc_serv fields.
Signed-off-by: NNeil Brown <neilb@suse.de>
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

WRITEMEM zeroes the last word in the destination buffer
for padding purposes, but this must not be done if
no bytes are to be copied, as it would result
in zeroing of the word right before the array.

The current implementation works since it's always called
with non zero nbytes or it follows an encoding of the
string (or opaque) length which, if equal to zero,
can be overwritten with zero.

Nevertheless, it seems safer to check for this case.
Signed-off-by: NBenny Halevy <bhalevy@panasas.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

We already print each operation of the compound when debugging is turned
on; printing the result could also help with remote debugging.
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

We're currently dereferencing the client after we drop our reference
count to it.
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

These bit operations don't need to be atomic.  They're all done under a
single big mutex anyway.
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

__FUNCTION__ is gcc-specific, use __func__
Signed-off-by: NHarvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Use proc_create() to make sure that ->proc_fops be setup before gluing PDE to
main tree.
Signed-off-by: NDenis V. Lunev <den@openvz.org>
Cc: Neil Brown <neilb@suse.de>
Cc: "J. Bruce Fields" <bfields@fieldses.org>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Presumably this is left over from earlier drafts of v4, which listed
TIME_METADATA as writeable.  It's read-only in rfc 3530, and shouldn't
be modifiable anyway.
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

The file_lock structure is used both as a heavy-weight representation of
an active lock, with pointers to reference-counted structures, etc., and
as a simple container for parameters that describe a file lock.

The conflicting lock returned from __posix_lock_file is an example of
the latter; so don't call the filesystem or lock manager callbacks when
copying to it.  This also saves the need for an unnecessary
locks_init_lock in the nfsv4 server.

Thanks to Trond for pointing out the error.
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>
Cc: Trond Myklebust <Trond.Myklebust@netapp.com>

Add /proc/fs/nfsd/unlock_filesystem, which allows e.g.:

shell> echo /mnt/sfs1 > /proc/fs/nfsd/unlock_filesystem

so that a filesystem can be unmounted before allowing a peer nfsd to
take over nfs service for the filesystem.
Signed-off-by: NS. Wendy Cheng <wcheng@redhat.com>
Cc: Lon Hohberger  <lhh@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

 fs/lockd/svcsubs.c          |   66 +++++++++++++++++++++++++++++++++++++++-----
 fs/nfsd/nfsctl.c            |   65 +++++++++++++++++++++++++++++++++++++++++++
 include/linux/lockd/lockd.h |    7 ++++
 3 files changed, 131 insertions(+), 7 deletions(-)

For high-availability NFS service, we generally need to be able to drop
file locks held on the exported filesystem before moving clients to a
new server.  Currently the only way to do that is by shutting down lockd
entirely, which is often undesireable (for example, if you want to
continue exporting other filesystems).

This patch allows the administrator to release all locks held by clients
accessing the client through a given server ip address, by echoing that
address to a new file, /proc/fs/nfsd/unlock_ip, as in:

shell> echo 10.1.1.2 > /proc/fs/nfsd/unlock_ip

The expected sequence of events can be:
1. Tear down the IP address
2. Unexport the path
3. Write IP to /proc/fs/nfsd/unlock_ip to unlock files
4. Signal peer to begin take-over.

For now we only support IPv4 addresses and NFSv2/v3 (NFSv4 locks are not
affected).

Also, if unmounting the filesystem is required, we assume at step 3 that
clients using the given server ip are the only clients holding locks on
the given filesystem; otherwise, an additional patch is required to
allow revoking all locks held by lockd on a given filesystem.
Signed-off-by: NS. Wendy Cheng <wcheng@redhat.com>
Cc: Lon Hohberger  <lhh@redhat.com>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

 fs/lockd/svcsubs.c          |   66 +++++++++++++++++++++++++++++++++++++++-----
 fs/nfsd/nfsctl.c            |   65 +++++++++++++++++++++++++++++++++++++++++++
 include/linux/lockd/lockd.h |    7 ++++
 3 files changed, 131 insertions(+), 7 deletions(-)

Currently, knfsd only clears the setuid bit if the owner of a file is
changed on a SETATTR call, and only clears the setgid bit if the group
is changed. POSIX says this in the spec for chown():

    "If the specified file is a regular file, one or more of the
     S_IXUSR, S_IXGRP, or S_IXOTH bits of the file mode are set, and the
     process does not have appropriate privileges, the set-user-ID
     (S_ISUID) and set-group-ID (S_ISGID) bits of the file mode shall
     be cleared upon successful return from chown()."

If I'm reading this correctly, then knfsd is doing this wrong. It should
be clearing both the setuid and setgid bit on any SETATTR that changes
the uid or gid. This wasn't really as noticable before, but now that the
ATTR_KILL_S*ID bits are a no-op for the NFS client, it's more evident.

This patch corrects the nfsd_setattr logic so that this occurs. It also
does a bit of cleanup to the function.

There is also one small behavioral change. If a SETATTR call comes in
that changes the uid/gid and the mode, then we now only clear the setgid
bit if the group execute bit isn't set. The setgid bit without a group
execute bit signifies mandatory locking and we likely don't want to
clear the bit in that case. Since there is no call in POSIX that should
generate a SETATTR call like this, then this should rarely happen, but
it's worth noting.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

...it's not really needed.
Signed-off-by: NJeff Layton <jlayton@redhat.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

There's no need to dynamically allocate this memory, and doing so may
create the possibility of races on shutdown of the rpc client.  (We've
witnessed it only after adding rpcsec_gss support to the server, after
which the rpc code can send destroys calls that expect to still be able
to access the rpc_stats structure after it has been destroyed.)

Such races are in theory possible if the module containing this "static"
memory is removed very quickly after an rpc client is destroyed, but
we haven't seen that happen.
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Move the code that actually parses the filehandle and looks up the
dentry and export to a separate function.  This simplifies the reference
counting a little and moves fh_verify() a little closer to the kernel
ideal of small, minimally-indentended functions.  Clean up a few other
minor style sins along the way.
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>
Cc: Neil Brown <neilb@suse.de>

While lease is correctly checked by supplying the type argument to
vfs_setlease(), it's stored with fl_type uninitialized. This breaks the
logic when checking the type of the lease.  The fix is to initialize
fl_type.

The old code still happened to function correctly since F_RDLCK is zero,
and we only implement read delegations currently (nor write
delegations).  But that's no excuse for not fixing this.
Signed-off-by: NFelix Blyakher <felixb@sgi.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

fs/nfsd/vfs.c:991:27: warning: Using plain integer as NULL pointer
Signed-off-by: NHarvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Add extern to nfsd/nfsd.h
fs/nfsd/nfssvc.c:146:5: warning: symbol 'nfsd_nrthreads' was not declared. Should it be static?
fs/nfsd/nfssvc.c:261:5: warning: symbol 'nfsd_nrpools' was not declared. Should it be static?
fs/nfsd/nfssvc.c:269:5: warning: symbol 'nfsd_get_nrthreads' was not declared. Should it be static?
fs/nfsd/nfssvc.c:281:5: warning: symbol 'nfsd_set_nrthreads' was not declared. Should it be static?
fs/nfsd/export.c:1534:23: warning: symbol 'nfs_exports_op' was not declared. Should it be static?

Add include of auth.h
fs/nfsd/auth.c:27:5: warning: symbol 'nfsd_setuser' was not declared. Should it be static?

Make static, move forward declaration closer to where it's needed.
fs/nfsd/nfs4state.c:1877:1: warning: symbol 'laundromat_main' was not declared. Should it be static?

Make static, forward declaration was already marked static.
fs/nfsd/nfs4idmap.c:206:1: warning: symbol 'idtoname_parse' was not declared. Should it be static?
fs/nfsd/vfs.c:1156:1: warning: symbol 'nfsd_create_setattr' was not declared. Should it be static?
Signed-off-by: NHarvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

This patch makes the needlessly global nfsd_create_setattr() static.
Signed-off-by: NAdrian Bunk <bunk@kernel.org>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

Likewise, distros usually leave CONFIG_NFSD_TCP enabled.

TCP support in the Linux NFS server is stable enough that we can leave it
on always.  CONFIG_NFSD_TCP adds about 10 lines of code, and defaults to
"Y" anyway.
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>

The code here is difficult to understand; attempt to clarify somewhat by
pulling out one of the more mystifying conditionals into a separate
function.

While we're here, also add lease_time to the list of attributes that we
don't really need to cross a mountpoint to fetch.
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>
Cc: Peter Staubach <staubach@redhat.com>

This condition is always true.
Signed-off-by: NJ. Bruce Fields <bfields@citi.umich.edu>