nfsd: Clean up errors in nfs3proc.c

Fix the following errors reported by checkpatch:

ERROR: need consistent spacing around '+' (ctx:WxV)
ERROR: spaces required around that '?' (ctx:VxW)

Signed-off-by: KaiLong Wang <wangkailong@jari.cn>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

nfsd: Clean up errors in nfs4state.c

Fix the following errors reported by checkpatch:

ERROR: spaces required around that '=' (ctx:VxW)

ERROR: space required after that ',' (ctx:VxO)
ERROR: space required before that '~' (ctx:OxV)
Signed-off-by: KaiLong Wang <wangkailong@jari.cn>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up errors in stats.c

Fix the following errors reported by checkpatch:

ERROR: space required after that ',' (ctx:VxV)

Signed-off-by: KaiLong Wang <wangkailong@jari.cn>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: simplify error paths in nfsd_svc()

The error paths in nfsd_svc() are needlessly complex and can result in a
final call to svc_put() without nfsd_last_thread() being called.  This
results in the listening sockets not being closed properly.

The per-netns setup provided by nfsd_startup_new() and removed by
nfsd_shutdown_net() is needed precisely when there are running threads.
So we don't need nfsd_up_before.  We don't need to know if it *was* up.
We only need to know if any threads are left.  If none are, then we must
call nfsd_shutdown_net().  But we don't need to do that explicitly as
nfsd_last_thread() does that for us.

So simply call nfsd_last_thread() before the last svc_put() if there are
no running threads.  That will always do the right thing.

Also discard:
 pr_info("nfsd: last server has exited, flushing export cache\n");
It may not be true if an attempt to start the first server failed, and
it isn't particularly helpful and it simply reports normal behaviour.

Signed-off-by: NeilBrown <neilb@suse.de>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up nfsd4_encode_seek()

Use modern XDR encoder utilities.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up nfsd4_encode_offset_status()

Use modern XDR encoder utilities.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up nfsd4_encode_copy_notify()

Replace open-coded encoding logic with the use of conventional XDR
utility functions.

Note that if we replace the cpn_sec and cpn_nsec fields with a
single struct timespec64 field, the encoder can use
nfsd4_encode_nfstime4(), as that is the data type specified by the
XDR spec.

NFS4ERR_INVAL seems inappropriate if the encoder doesn't support
encoding the response. Instead use NFS4ERR_SERVERFAULT, since this
condition is a software bug on the server.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up nfsd4_encode_copy()

Restructure this function using conventional XDR utility functions
and so it aligns better with the XDR in the specification.

I've also moved nfsd4_encode_copy() closer to the data type encoders
that only it uses.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up nfsd4_encode_test_stateid()

Use conventional XDR utilities.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up nfsd4_encode_exchange_id()

Restructure nfsd4_encode_exchange_id() so that it will be more
straightforward to add support for SSV one day. Also, adopt the use
of the conventional XDR utility functions.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up nfsd4_do_encode_secinfo()

Refactor nfsd4_encode_secinfo() so it is more clear what XDR data
item is being encoded by which piece of code.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up nfsd4_encode_access()

Convert nfsd4_encode_access() to use modern XDR utility functions.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up nfsd4_encode_readdir()

Untangle nfsd4_encode_readdir() so it is more clear what XDR data
item is being encoded by which piece of code.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up nfsd4_encode_entry4()

Reshape nfsd4_encode_entry4() to be more like the legacy dirent
encoders, which were recently rewritten to use xdr_stream.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add an nfsd4_encode_nfs_cookie4() helper

De-duplicate the entry4 cookie encoder, similar to the arrangement
for the NFSv2 and NFSv3 directory entry encoders.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up nfsd4_encode_rdattr_error()

No need for specialized code here, as this function is invoked only
rarely. Convert it to encode to xdr_stream using conventional XDR
helpers.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Rename nfsd4_encode_dirent()

Rename nfsd4_encode_dirent() to match the naming convention already
used in the NFSv2 and NFSv3 readdir paths. The new name reflects the
name of the spec-defined XDR data type for an NFSv4 directory entry.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up nfsd4_encode_sequence()

De-duplicate open-coded encoding of the sessionid, and convert the
rest of the function to use conventional XDR utility functions.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Restructure nfsd4_encode_create_session()

Convert nfsd4_encode_create_session() to use the conventional XDR
encoding utilities.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_channel_attr4()

De-duplicate the encoding of the fore channel and backchannel
attributes.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add a utility function for encoding sessionid4 objects

There is more than one NFSv4 operation that needs to encode a
sessionid4, so extract that data type into a separate helper.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up nfsd4_encode_open()

Finish cleaning up nfsd4_encode_open().

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_open_delegation4()

To better align our implementation with the XDR specification,
refactor the part of nfsd4_encode_open() that encodes delegation
metadata.

As part of that refactor, remove an unnecessary BUG() call site and
a comment that appears to be stale.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_open_none_delegation4()

To better align our implementation with the XDR specification,
refactor the part of nfsd4_encode_open() that encodes the
open_none_delegation4 type.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_open_write_delegation4()

Make it easier to adjust the XDR encoder to handle new features
related to write delegations.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_open_read_delegation4()

Refactor nfsd4_encode_open() so the open_read_delegation4 type is
encoded in a separate function. This makes it more straightforward
to later add support for returning an nfsace4 in OPEN responses that
offer a delegation.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Refactor nfsd4_encode_lock_denied()

Use the modern XDR utility functions.

The LOCK and LOCKT encoder functions need to return nfserr_denied
when a lock is denied, but nfsd4_encode_lock4denied() should return
a status code that is consistent with other XDR encoders.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_lock_owner4()

To improve readability and better align the LOCK encoders with the
XDR specification, add an explicit encoder named for the lock_owner4
type.

In particular, to avoid code duplication, use
nfsd4_encode_clientid4() to encode the clientid in the lock owner
rather than open-coding it.

It looks to me like nfs4_set_lock_denied() already clears the
clientid if it won't return an owner (cf: the nevermind: label). The
code in the XDR encoder appears to be redundant and can safely be
removed.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Remove a layering violation when encoding lock_denied

An XDR encoder is responsible for marshaling results, not releasing
memory that was allocated by the upper layer. We have .op_release
for that purpose.

Move the release of the ld_owner.data string to op_release functions
for LOCK and LOCKT.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up nfsd4_encode_getdeviceinfo()

Adopt the conventional XDR utility functions. Also, restructure to
make the function align more closely with the spec -- there doesn't
seem to be a performance need for speciality code, so prioritize
readability.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Make @gdev parameter of ->encode_getdeviceinfo a const pointer

This enables callers to be passed const pointer parameters.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up nfsd4_encode_layoutreturn()

Adopt the use of conventional XDR utility functions. Restructure
the encoder to better align with the XDR definition of the result.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up nfsd4_encode_layoutcommit()

Adopt the use of conventional XDR utility functions. Restructure
the encoder to better align with the XDR definition of the result.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up nfsd4_encode_layoutget()

De-duplicate the open-coded stateid4 encoder. Adopt the use of the
conventional current XDR encoding helpers. Refactor the encoder to
align with the XDR specification.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Make @lgp parameter of ->encode_layoutget a const pointer

This enables callers to be passed const pointer parameters.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up nfsd4_encode_stateid()

Update the encoder function name to match the type name, as is the
convention with other such encoder utility functions, and with
nfsd4_decode_stateid4().

Make the @stateid argument a const so that callers of
nfsd4_encode_stateid4() in the future can be passed const pointers
to structures.

Since the compiler is allowed to add padding to structs, use the
wire (spec-defined) size when reserving buffer space.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_count4()

This is a synonym for nfsd4_encode_uint32_t() that matches the
name of the XDR type. It will get at least one more use in a
subsequent patch.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Rename nfsd4_encode_fattr()

For better alignment with the specification, NFSD's encoder function
name should match the name of the XDR data type.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Use a bitmask loop to encode FATTR4 results

The fattr4 encoder is now structured like the COMPOUND op encoder:
one function for each individual attribute, called by bit number.
Benefits include:

- The individual attributes are now guaranteed to be encoded in
  bitmask order into the send buffer

- There can be no unwanted side effects between attribute encoders

- The code now clearly documents which attributes are /not/
  implemented on this server

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Copy FATTR4 bit number definitions from RFCs

I'd like to convert nfsd4_encode_fattr() to rotate through the
attrmask using for_each_bit() instead of explicitly testing the
bitmask for each bit value. This means I need the bit numbers, as
defined in the specs, instead of our internal bitmask constants.

As a clean up, use the new spec-derived values to define the WORD#
bitmask constants.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_xattr_support()

Refactor the encoder for FATTR4_XATTR_SUPPORT into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_sec_label()

Refactor the encoder for FATTR4_SEC_LABEL into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_suppattr_exclcreat()

Refactor the encoder for FATTR4_SUPPATTR_EXCLCREAT into a helper. In
a subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_layout_blksize()

Refactor the encoder for FATTR4_LAYOUT_BLKSIZE into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_layout_types()

Refactor the encoder for FATTR4_LAYOUT_TYPES into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_fs_layout_types()

Refactor the encoder for FATTR4_FS_LAYOUT_TYPES into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_mounted_on_fileid()

Refactor the encoder for FATTR4_MOUNTED_ON_FILEID into a helper. In
a subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_time_modify()

Refactor the encoder for FATTR4_TIME_MODIFY into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_time_metadata()

Refactor the encoder for FATTR4_TIME_METADATA into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_time_delta()

Refactor the encoder for FATTR4_TIME_DELTA into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

fattr4_time_delta is specified as an nfstime4, so de-duplicate this
encoder.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_time_create()

Refactor the encoder for FATTR4_TIME_CREATE into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_time_access()

Refactor the encoder for FATTR4_TIME_ACCESS into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_space_used()

Refactor the encoder for FATTR4_SPACE_USED into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_space_total()

Refactor the encoder for FATTR4_SPACE_TOTAL into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_space_free()

Refactor the encoder for FATTR4_SPACE_FREE into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_space_avail()

Refactor the encoder for FATTR4_SPACE_AVAIL into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_rawdev()

Refactor the encoder for FATTR4_RAWDEV into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_owner_group()

Refactor the encoder for FATTR4_OWNER_GROUP into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_owner()

Refactor the encoder for FATTR4_OWNER into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_numlinks()

Refactor the encoder for FATTR4_NUMLINKS into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_mode()

Refactor the encoder for FATTR4_MODE into a helper. In a subsequent
patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_maxwrite()

Refactor the encoder for FATTR4_MAXWRITE into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_maxread()

Refactor the encoder for FATTR4_MAXREAD into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_maxname()

Refactor the encoder for FATTR4_MAXNAME into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_maxlink()

Refactor the encoder for FATTR4_MAXLINK into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_maxfilesize()

Refactor the encoder for FATTR4_MAXFILESIZE into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_fs_locations()

Refactor the encoder for FATTR4_FS_LOCATIONS into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_files_total()

Refactor the encoder for FATTR4_FILES_TOTAL into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_files_free()

Refactor the encoder for FATTR4_FILES_FREE into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_files_avail()

Refactor the encoder for FATTR4_FILES_AVAIL into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_fileid()

Refactor the encoder for FATTR4_FILEID into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_filehandle()

Refactor the encoder for FATTR4_FILEHANDLE into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

We can de-duplicate the other filehandle encoder (in GETFH) using
our new helper.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_acl()

Refactor the encoder for FATTR4_ACL into a helper. In a subsequent
patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_nfsace4()

Refactor the ACE encoding helper so that it can eventually be reused
for encoding OPEN results that contain delegation ACEs.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_aclsupport()

Refactor the encoder for FATTR4_ACLSUPPORT into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_rdattr_error()

Refactor the encoder for FATTR4_RDATTR_ERROR into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_lease_time()

Refactor the encoder for FATTR4_LEASE_TIME into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_fsid()

Refactor the encoder for FATTR4_FSID into a helper. In a subsequent
patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_size()

Refactor the encoder for FATTR4_SIZE into a helper. In a subsequent
patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_change()

Refactor the encoder for FATTR4_CHANGE into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

The code is restructured a bit to use the modern xdr_stream flow,
and the encoded cinfo value is made const so that callers of the
encoders can be passed a const cinfo.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_fh_expire_type()

Refactor the encoder for FATTR4_FH_EXPIRE_TYPE into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_type()

Refactor the encoder for FATTR4_TYPE into a helper. In a subsequent
patch, this helper will be called from a bitmask loop.

In addition, restructure the code so that byte-swapping is done on
constant values rather than at run time. Run-time swapping can be
costly on some platforms, and "type" is a frequently-requested
attribute.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4_supported_attrs()

Refactor the encoder for FATTR4_SUPPORTED_ATTRS into a helper. In a
subsequent patch, this helper will be called from a bitmask loop.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4__false()

Add an encoding helper that encodes a single boolean "false" value.
Attributes that always return "false" can use this helper.

In a subsequent patch, this helper will be called from a bitmask
loop, so it is given a standardized synopsis.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add nfsd4_encode_fattr4__true()

Add an encoding helper that encodes a single boolean "true" value.
Attributes that always return "true" can use this helper.

In a subsequent patch, this helper will be called from a bitmask
loop, so it is given a standardized synopsis.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add struct nfsd4_fattr_args

I'm about to split nfsd4_encode_fattr() into a number of smaller
functions. Instead of passing a large number of arguments to each of
the smaller functions, create a struct that can gather the common
argument variables into something with a convenient handle on it.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Clean up nfsd4_encode_setattr()

De-duplicate the encoding of bitmap4 results in
nfsd4_encode_setattr().

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Rename nfsd4_encode_bitmap()

For alignment with the specification, the name of NFSD's encoder
function should match the name of the XDR type.

I've also replaced a few "naked integers" with symbolic constants
that better reflect the usage of these values.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: Add simple u32, u64, and bool encoders

The generic XDR encoders return a length or a negative errno. NFSv4
encoders want to know simply whether the encode ran out of stream
buffer space. The return values for server-side encoding are either
nfs_ok or nfserr_resource.

So far I've found it adds a lot of duplicate code to try to use the
generic XDR encoder utilities when encoding the simple data types in
the NFSv4 operation encoders.

Add a set of NFSv4-specific utilities that handle the basic XDR data
types. These are added in xdr4.h so they might eventually be used by
the callback server and pNFS driver encoders too.

Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

SUNRPC: Remove BUG_ON call sites

There is no need to take down the whole system for these assertions.

I'd rather not attempt a heroic save here, as some bug has occurred
that has left the transport data structures in an unknown state.
Just warn and then leak the left-over resources.

Acked-by: Christian Brauner <brauner@kernel.org>
Reviewed-by: NeilBrown <neilb@suse.de>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

nfs: fix the typo of rfc number about xattr in NFSv4

Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

tools: ynl: Add source files for nfsd netlink protocol

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: add rpc_status netlink support

Introduce rpc_status netlink support for NFSD in order to dump pending
RPC requests debugging information from userspace.

Closes: https://bugzilla.linux-nfs.org/show_bug.cgi?id=366
Tested-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: introduce netlink stubs

Generate stubs and uAPI for nfsd netlink protocol. For the moment,
the new protocol has one operation: rpc_status.

The generated header and source files are created by running:

  tools/net/ynl/ynl-regen.sh

Tested-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Acked-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: handle GETATTR conflict with write delegation

If the GETATTR request on a file that has write delegation in effect
and the request attributes include the change info and size attribute
then the request is handled as below:

Server sends CB_GETATTR to client to get the latest change info and file
size. If these values are the same as the server's cached values then
the GETATTR proceeds as normal.

If either the change info or file size is different from the server's
cached values, or the file was already marked as modified, then:

    . update time_modify and time_metadata into file's metadata
      with current time

    . encode GETATTR as normal except the file size is encoded with
      the value returned from CB_GETATTR

    . mark the file as modified

If the CB_GETATTR fails for any reasons, the delegation is recalled
and NFS4ERR_DELAY is returned for the GETATTR.

Signed-off-by: Dai Ngo <dai.ngo@oracle.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

NFSD: add support for CB_GETATTR callback

Includes:
   . CB_GETATTR proc for nfs4_cb_procedures[]
   . XDR encoding and decoding function for CB_GETATTR request/reply
   . add nfs4_cb_fattr to nfs4_delegation for sending CB_GETATTR
     and store file attributes from client's reply.

Signed-off-by: Dai Ngo <dai.ngo@oracle.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

SUNRPC: change the back-channel queue to lwq

This removes the need to store and update back-links in the list.
It also remove the need for the _bh version of spin_lock().

Signed-off-by: NeilBrown <neilb@suse.de>
Cc: Trond Myklebust <trond.myklebust@hammerspace.com>
Cc: Anna Schumaker <Anna.Schumaker@Netapp.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

SUNRPC: discard sp_lock

sp_lock is now only used to protect sp_all_threads.  This isn't needed
as sp_all_threads is only manipulated through svc_set_num_threads(),
which is already serialized.  Read-acccess only requires rcu_read_lock().
So no more locking is needed.

Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

SUNRPC: change sp_nrthreads to atomic_t

Using an atomic_t avoids the need to take a spinlock (which can soon be
removed).

Choosing a thread to kill needs to be careful as we cannot set the "die
now" bit atomically with the test on the count.  Instead we temporarily
increase the count.

Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>

SUNRPC: use lwq for sp_sockets - renamed to sp_xprts

lwq avoids using back pointers in lists, and uses less locking.
This introduces a new spinlock, but the other one will be removed in a
future patch.

For svc_clean_up_xprts(), we now dequeue the entire queue, walk it to
remove and process the xprts that need cleaning up, then re-enqueue the
remaining queue.

Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>