Right now, we can call nfs_commit_inode() while holding the session slot,
which could lead to NFSv4 deadlocks. Ensure we only keep the slot if
the server returned a layout that we have to process.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

I was able to reproduce this pretty regularily using xfstests
generic/013 on NFS v4.0.
Reported-by: NRoss Zwisler <Ross.Zwisler@linux.intel.com>
Fixes: 6c342655 (NFSv4: Return NFS4ERR_DELAY when a delegation recall fails due to igrab())
Signed-off-by: NAnna Schumaker <Anna.Schumaker@Netapp.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Currently, when IO to DS fails, client returns the layout and
retries against the MDS. However, then on umounting (inode eviction)
it returns the layout again.

This is because pnfs_return_layout() was changed in
commit d78471d3 ("pnfs/blocklayout: set PNFS_LAYOUTRETURN_ON_ERROR")
to always set NFS_LAYOUT_RETURN_REQUESTED so even if we returned
the layout, it will be returned again. Instead, let's also check
if we have already marked the layout invalid.
Signed-off-by: NOlga Kornievskaia <kolga@netapp.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

If the server returns NFS4ERR_SEQ_FALSE_RETRY or NFS4ERR_RETRY_UNCACHED_REP,
then it thinks we're trying to replay an existing request. If so, then
let's just bump the sequence ID and retry the operation.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

We want to compare the slot_id to the highest slot number advertised by the
server.

Fixes: 3be0f80b ("NFSv4.1: Fix up replays of interrupted requests")
Cc: stable@vger.kernel.org # 4.15+
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

The correct behaviour for NFSv4 sequence IDs is to wrap around
to the value 0 after 0xffffffff.
See https://tools.ietf.org/html/rfc5661#section-2.10.6.1

Fixes: 5f83d86c ("NFSv4.x: Fix wraparound issues when validing...")
Cc: stable@vger.kernel.org # 4.6+
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

If the attempt to recall the delegation fails because the inode is
in the process of being evicted from cache, then use NFS4ERR_DELAY
to ask the server to retry later.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Commit 69dd716c ("NFSv4: Add socket proto argument to
setclientid") (2007) added the transport protocol name to the client
ID string, but the patch description doesn't explain why this was
necessary.

At that time, the only transport protocol name that would have been
used is "tcp" (for both IPv4 and IPv6), resulting in no additional
distinctiveness of the client ID string.

Since there is one client instance, the server should recognize it's
state whether the client is connecting via TCP or RDMA. Same client,
same lease.
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

It is possible for two distinct clients to have the same cl_ipaddr:

 - if the client admin disables callback with clientaddr=0.0.0.0 on
   more than one client

 - if two clients behind separate NATs use the same private subnet
   number

 - if the client admin specifies the same address via clientaddr=
   mount option (pointing the server at the same NAT box, for
   example)

Because of the way the Linux NFSv4.0 client constructs its client
ID string by default, such clients could interfere with each others'
lease state when mounting the same server:

	scnprintf(str, len, "Linux NFSv4.0 %s/%s %s",
		clp->cl_ipaddr,
		rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR),
		rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_PROTO));

cl_ipaddr is set to the value of the clientaddr= mount option. Two
clients whose addresses are 192.168.3.77 that mount the same server
(whose public IP address is, say, 3.4.5.6) would both generate the
same client ID string when sending a SETCLIENTID:

  Linux NFSv4.0 192.168.3.77/3.4.5.6 tcp

and thus the server would not be able to distinguish the clients'
leases. If both clients are using AUTH_SYS when sending SETCLIENTID
then the server could possibly permit the two clients to interfere
with or purge each others' leases.

To better ensure that Linux's NFSv4.0 client ID strings are distinct
in these cases, remove cl_ipaddr from the client ID string and
replace it with something more likely to be unique. Note that the
replacement looks a lot like the uniform client ID string.
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Fix a compiler warning:
fs/nfs/nfs4proc.c:910:13: warning: 'nfs4_layoutget_release' defined but not used [-Wunused-function]
 static void nfs4_layoutget_release(void *calldata)
             ^~~~~~~~~~~~~~~~~~~~~~
Reported-by: NStephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

NFS-over-RDMA client updates for Linux 4.18

Stable patches:
- xprtrdma: Return -ENOBUFS when no pages are available

New features:
- Add ->alloc_slot() and ->free_slot() functions

Bugfixes and cleanups:
- Add missing SPDX tags to some files
- Try to fail mount quickly if client has no RDMA devices
- Create transport IDs in the correct network namespace
- Fix max_send_wr computation
- Clean up receive tracepoints
- Refactor receive handling
- Remove unused functions

If the client holds a delegation, then ensure we filter out attempts
to invalidate the size, owner, group owner, or mode unless we made the
change, in which case, check that NFS_INO_REVAL_FORCED is set by the
caller.
Always filter out attempts to invalidate the change attribute and
size, since we are authoritative for those.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

If we hold a delegation, we should not need to call
nfs_check_inode_attributes() since we already know which attributes
are valid, and which ones may still need revalidation. The state
of the NFS_INO_REVAL_FORCED flag is therefore irrelevant.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Make sure that the client completely ignores change attribute and size
changes on the server when it holds a delegation.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Don't mark attributes as invalid just because they have changed. Instead,
for the purposes of adjusting the attribute cache timeout, keep a
separate variable that tracks whether or not a change occurred.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Always try to set the attributes, even if we don't have a valid struct
nfs_fattr.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

If there are attributes that are still invalid when we set a delegation,
then we need to set the NFS_INO_REVAL_FORCED flag.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

If we hold a delegation, we don't need to care about whether or not
the inode attributes are up to date. We know we can cache the results
of this call regardless.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Again, when revalidating the inode, we don't need to ask for attributes
for which we are authoritative.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Allow the getattr() callback to check things like whether or not we hold
a delegation so that it can adjust the attributes that it is asking for.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

When we hold a delegation, we should not need to request attributes such
as the file size or the change attribute. For some servers, avoiding
asking for these unneeded attributes can improve the overall system
performance.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Clean up: This array was used in a dprintk that was replaced by a
trace point in commit ab03eff5 ("xprtrdma: Add trace points in
RPC Call transmit paths").
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NAnna Schumaker <Anna.Schumaker@Netapp.com>

Matches trace_xprtrdma_dma_unmap(mr).
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NAnna Schumaker <Anna.Schumaker@Netapp.com>

Currently, when the sendctx queue is exhausted during marshaling, the
RPC/RDMA transport places the RPC task on the delayq, which forces a
wait for HZ >> 2 before the marshal and send is retried.

With this change, the transport now places such an RPC task on the
pending queue, and wakes it just as soon as more sendctxs become
available. This typically takes less than a millisecond, and the
write_space waking mechanism is less deadlock-prone.

Moreover, the waiting RPC task is holding the transport's write
lock, which blocks the transport from sending RPCs. Therefore faster
recovery from sendctx queue exhaustion is desirable.

Cf. commit 5804891455d5 ("xprtrdma: ->send_request returns -EAGAIN
when there are no free MRs").
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NAnna Schumaker <Anna.Schumaker@Netapp.com>

Clean up: The logic to wait for write space is common to a bunch of
the encoding helper functions. Lift it out and put it in the tail
of rpcrdma_marshal_req().
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Signed-off-by: NAnna Schumaker <Anna.Schumaker@Netapp.com>

The use of -EAGAIN in rpcrdma_convert_iovs() is a latent bug: the
transport never calls xprt_write_space() when more pages become
available. -ENOBUFS will trigger the correct "delay briefly and call
again" logic.

Fixes: 7a89f9c6 ("xprtrdma: Honor ->send_request API contract")
Signed-off-by: NChuck Lever <chuck.lever@oracle.com>
Cc: stable@vger.kernel.org # 4.8+
Signed-off-by: NAnna Schumaker <Anna.Schumaker@Netapp.com>

If the server recalls the layout that was just handed out, we risk hitting
a race as described in RFC5661 Section 2.10.6.3 unless we ensure that we
release the sequence slot after processing the LAYOUTGET operation that
was sent as part of the OPEN compound.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

If the layoutget on open call failed, we can't really commit the inode,
so don't bother calling it.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

If we're only opening the file for reading, and the file is empty and/or
we already have cached data, then heuristically optimise away the
LAYOUTGET.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Ensure that we only switch off the LAYOUTGET operation in the OPEN
compound when the server is truly broken, and/or it is complaining
that the compound is too large.
Currently, we end up turning off the functionality permanently,
even for transient errors such as EACCES or ENOSPC.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

We need to ensure that pnfs_parse_lgopen() doesn't try to parse a
struct nfs4_layoutget_res that was not filled by a successful call
to decode_layoutget(). This can happen if we performed a cached open,
or if either the OP_ACCESS or OP_GETATTR operations preceding the
OP_LAYOUTGET in the compound returned an error.

By initialising the 'status' field to NFS4ERR_DELAY, we ensure that
pnfs_parse_lgopen() won't try to interpret the structure.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

The flag was not always being cleared after LAYOUTGET on OPEN.
Signed-off-by: NFred Isaman <fred.isaman@gmail.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Since the LAYOUTGET on OPEN can be sent without prior inode information,
existing methods to prevent LAYOUTGET from being sent while processing
CB_LAYOUTRECALL don't work.  Track if a recall occurred while LAYOUTGET
was being sent, and if so ignore the results.
Signed-off-by: NFred Isaman <fred.isaman@gmail.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Signed-off-by: NFred Isaman <fred.isaman@gmail.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Signed-off-by: NFred Isaman <fred.isaman@gmail.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Move the actual freeing of the struct nfs4_layoutget into fs/nfs/pnfs.c
where it can be reused by the layoutget on open code.
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

This triggers when have no pre-existing inode to attach to.
The preexisting case is saved for later.
Signed-off-by: NFred Isaman <fred.isaman@gmail.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>

Don't send in a layout, instead use the (possibly NULL) inode.

This is needed for LAYOUTGET attached to an OPEN where the inode is not
yet set.
Signed-off-by: NFred Isaman <fred.isaman@gmail.com>
Signed-off-by: NTrond Myklebust <trond.myklebust@hammerspace.com>