Tools like tcpdump and rpcdebug can be very useful. But there are
plenty of environments where they are difficult or impossible to
use. For example, we've had customers report I/O failures during
workloads so heavy that collecting network traffic or enabling
RPC debugging are themselves onerous.

The kernel's static tracepoints are lightweight (less likely to
introduce timing changes) and efficient (the trace data is compact).
They also work in scenarios where capturing network traffic is not
possible due to lack of hardware support (some InfiniBand HCAs) or
where data or network privacy is a concern.

Introduce tracepoints that show when an NFS READ, WRITE, or COMMIT
is initiated, and when it completes. Record the arguments and
results of each operation, which are not shown by existing sunrpc
module's tracepoints.

For instance, the recorded offset and count can be used to match an
"initiate" event to a "done" event. If an NFS READ result returns
fewer bytes than requested or zero, seeing the EOF flag can be
probative. Seeing an NFS4ERR_BAD_STATEID result is also indication
of a particular class of problems. The timing information attached
to each event record can often be useful as well.

Usage example:

[root@manet tmp]# trace-cmd record -e nfs:*initiate* -e nfs:*done
/sys/kernel/debug/tracing/events/nfs/*initiate*/filter
/sys/kernel/debug/tracing/events/nfs/*done/filter
Hit Ctrl^C to stop recording
^CKernel buffer statistics:
  Note: "entries" are the entries left in the kernel ring buffer and are not
        recorded in the trace data. They should all be zero.

CPU: 0
entries: 0
overrun: 0
commit overrun: 0
bytes: 3680
oldest event ts:    78.367422
now ts:   100.124419
dropped events: 0
read events: 74

... and so on.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Instead of having a private method for copying the open/delegation stateid,
use the same call that is used for standard I/O through the MDS.

Note that this means we transmit the stateid with a zero seqid, avoiding
issues with NFS4ERR_OLD_STATEID.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

If we skip a subrequest due to a zero refcount, we should still count
the byte range that it covered so that we accurately reconstruct the
original request size.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

That can deadlock if this is the last reference since
nfs_page_group_destroy() calls nfs_page_group_sync_on_bit().
Note that even if the page was removed from the subpage list,
the req->wb_head could still be pointing to the old head.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

It's pretty much a duplicate of nfs_scan_commit_list() that also
clears the PG_COMMIT_TO_DS flag.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Since the commit list is not ordered, it is possible for nfs_scan_commit_list
to hold a request that nfs_lock_and_join_requests() is waiting for, while
at the same time trying to grab a request that nfs_lock_and_join_requests
already holds.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

The writeback code wants to send a commit after processing the pages,
which is why we want to delay releasing the struct path until after
that's done.

Also, the layout code expects that we do not free the inode before
we've put the layout segments in pnfs_writehdr_free() and
pnfs_readhdr_free()

Fixes: 919e3bd9 ("NFS: Ensure we commit after writeback is complete")
Fixes: 4714fb51 ("nfs: remove pgio_header refcount, related cleanup")
Cc: stable@vger.kernel.org
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Since commit 18290650 ("NFS: Move buffered I/O locking into
nfs_file_write()") nfs_file_write() has not flushed the correct byte
range during synchronous writes.  generic_write_sync() expects that
iocb->ki_pos points to the right edge of the range rather than the
left edge.

To replicate the problem, open a file with O_DSYNC, have the client
write at increasing offsets, and then print the successful offsets.
Block port 2049 partway through that sequence, and observe that the
client application indicates successful writes in advance of what the
server received.

Fixes: 18290650 ("NFS: Move buffered I/O locking into nfs_file_write()")
Signed-off-by: Jacob Strauss <jsstraus@amazon.com>
Signed-off-by: Tarang Gupta <tarangg@amazon.com>
Tested-by: Tarang Gupta <tarangg@amazon.com>
Cc: stable@vger.kernel.org # v4.8+
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Omit an extra message for a memory allocation failure in this function.

This issue was detected by using the Coccinelle software.
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

This field hasn't been used since commit 57b69181 ("NFS: Cache
access checks more aggressively").
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

When a byte range lock (or flock) is taken out on an NFS file, the
validity of the cached data is checked and the inode is marked
NFS_INODE_INVALID_DATA.  However the cached data isn't flushed from
the page cache.

This is sufficient for future read() requests or mmap() requests as
they call nfs_revalidate_mapping() which performs the flush if
necessary.

However an existing mapping is not affected.  Accessing data through
that mapping will continue to return old data even though the inode is
marked NFS_INODE_INVALID_DATA.

This can easily be confirmed using the 'nfs' tool in
  git://github.com/okirch/twopence-nfs.git
and running

   nfs coherence FILENAME
on one client, and
   nfs coherence -r FILENAME
on another client.

It appears that prior to Linux 2.6.0 this worked correctly.

However commit:

http://git.kernel.org/cgit/linux/kernel/git/history/history.git/commit/?id=ca9268fe3ddd075714005adecd4afbd7f9ab87d0

removed the call to inode_invalidate_pages() from nfs_zap_caches().  I
haven't tested this code, but inspection suggests that prior to this
commit, file locking would invalidate all inode pages.

This patch adds a call to nfs_revalidate_mapping() after a
successful SETLK so that invalid data is flushed.  With this patch the
above test passes.  To minimize impact (and possibly avoid a GETATTR
call) this only happens if the mapping might be mapped into
userspace.

Cc: Olaf Kirch <okir@suse.com>
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

RPC_TASK_NO_RETRANS_TIMEOUT is set when cl_noretranstimeo
is set, which happens when  RPC_CLNT_CREATE_NO_RETRANS_TIMEOUT is set,
which happens when NFS_CS_NO_RETRANS_TIMEOUT is set.

This flag means "don't resend on a timeout, only resend if the
connection gets broken for some reason".

cl_discrtry is set when RPC_CLNT_CREATE_DISCRTRY is set, which
happens when NFS_CS_DISCRTRY is set.

This flag means "always disconnect before resending".

NFS_CS_NO_RETRANS_TIMEOUT and NFS_CS_DISCRTRY are both only set
in nfs4_init_client(), and it always sets both.

So we will never have a situation where only one of the flags is set.
So this code, which tests if timeout retransmits are allowed, and
disconnection is required, will never run.

So it makes sense to remove this code as it cannot be tested and
could confuse people reading the code (like me).

(alternately we could leave it there with a comment saying
 it is never actually used).
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Commit fbe77c30 ("NFS: move rw_mode to nfs_pageio_header")
reintroduced some pointless code that commit 518662e0 ("NFS: fix
usage of mempools.") had recently removed.

Remove it again.

Cc: Benjamin Coddington <bcodding@redhat.com>
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Adopt the use of xprt_pin_rqst to eliminate contention between
Call-side users of rb_lock and the use of rb_lock in
rpcrdma_reply_handler.

This replaces the mechanism introduced in 431af645 ("xprtrdma:
Fix client lock-up after application signal fires").

Use recv_lock to quickly find the completing rqst, pin it, then
drop the lock. At that point invalidation and pull-up of the Reply
XDR can be done. Both are often expensive operations.

Finally, take recv_lock again to signal completion to the RPC
layer. It also protects adjustment of "cwnd".

This greatly reduces the amount of time a lock is held by the
reply handler. Comparing lock_stat results shows a marked decrease
in contention on rb_lock and recv_lock.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
[trond.myklebust@primarydata.com: Remove call to rpcrdma_buffer_put() from
   the "out_norqst:" path in rpcrdma_reply_handler.]
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

NFS-over-RDMA client updates for Linux 4.14

Bugfixes and cleanups:
- Constify rpc_xprt_ops
- Harden RPC call encoding and decoding
- Clean up rpc call decoding to use xdr_streams
- Remove unused variables from various structures
- Refactor code to remove imul instructions
- Rearrange rx_stats structure for better cacheline sharing

To reduce false cacheline sharing, separate counters that are likely
to be accessed in the Call path from those accessed in the Reply
path.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>

For a while now any NFSv2 mount where sec= is specified uses
AUTH_NULL. If sec= is not specified, the mount uses AUTH_UNIX.
Commit e68fd7c8 ("mount: use sec= that was specified on the
command line") attempted to address a very similar problem with
NFSv3, and should have fixed this too, but it has a bug.

The MNTv1 MNT procedure does not return a list of security flavors,
so our client makes up a list containing just AUTH_NULL. This should
enable nfs_verify_authflavors() to assign the sec= specified flavor,
but instead, it incorrectly sets it to AUTH_NULL.

I expect this would also be a problem for any NFSv3 server whose
MNTv3 MNT procedure returned a security flavor list containing only
AUTH_NULL.

Fixes: e68fd7c8 ("mount: use sec= that was specified on ... ")
BugLink: https://bugzilla.linux-nfs.org/show_bug.cgi?id=310Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

An NFSv4.1 client might close a file after the user who opened it has
logged off.  In this case the user's credentials may no longer be
valid, if they are e.g. kerberos credentials that have expired.

NFSv4.1 has a mechanism to allow the client to use machine credentials
to close a file.  However due to a short-coming in the RFC, a CLOSE
with those credentials may not be possible if the file in question
isn't exported to the same security flavor - the required PUTFH must
be rejected when this is the case.

Specifically if a server and client support kerberos in general and
have used it to form a machine credential, but the file is only
exported to "sec=sys", a PUTFH with the machine credentials will fail,
so CLOSE is not possible.

As RPC_AUTH_UNIX (used by sec=sys) credentials can never expire, there
is no value in using the machine credential in place of them.
So in that case, just use the users credentials for CLOSE etc, as you would
in NFSv4.0
Signed-off-by: Neil Brown <neilb@suse.com>
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

If you
 - mount and NFSv3 filesystem
 - do some file locking which requires the server
   to make a GRANT call back
 - unmount
 - mount again and do the same locking

then the second attempt at locking suffers a 30 second delay.
Unmounting and remounting causes lockd to stop and restart,
which causes it to bind to a new port.
The server still thinks the old port is valid and gets ECONNREFUSED
when trying to contact it.
ECONNREFUSED should be seen as a hard error that is not worth
retrying.  Rebinding is the only reasonable response.

This patch forces a rebind if that makes sense.
Signed-off-by: NeilBrown <neilb@suse.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Now that the mirror allocation has been moved, the parameter can go.
Also remove the redundant symbol export.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

There are a number of callers of nfs_pageio_complete() that want to
continue using the nfs_pageio_descriptor without needing to call
nfs_pageio_init() again. Examples include nfs_pageio_resend() and
nfs_pageio_cond_complete().

The problem is that nfs_pageio_complete() also calls
nfs_pageio_cleanup_mirroring(), which frees up the array of mirrors.
This can lead to writeback errors, in the next call to
nfs_pageio_setup_mirroring().

Fix by simply moving the allocation of the mirrors to
nfs_pageio_setup_mirroring().

Link: https://bugzilla.kernel.org/show_bug.cgi?id=196709Reported-by: JianhongYin <yin-jianhong@163.com>
Cc: stable@vger.kernel.org # 4.0+
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

This further reduces contention with the transport_lock, and allows us
to convert to using a non-bh-safe spinlock, since the list is now never
accessed from a bh context.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Simplify the code to avoid a full copy of the struct xdr_skb_reader.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Ensure that we don't hog the workqueue thread by requeuing the job
every 64 loops.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

The backchannel request has no associated task, so it is going nowhere
until we call xprt_complete_bc_request().
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Instead add a mechanism to ensure that the request doesn't disappear
from underneath us while copying from the socket. We do this by
preventing xprt_release() from freeing the XDR buffers until the
flag RPC_TASK_MSG_RECV has been cleared from the request.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Reviewed-by: Chuck Lever <chuck.lever@oracle.com>

Re-arrange the pointer arithmetic in the chunk list encoders to
eliminate several more integer multiplication instructions during
Transport Header encoding.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>

Re-arrange the pointer arithmetic in rpcrdma_convert_iovs() to
eliminate several integer multiplication instructions during
Transport Header encoding.

Also, array overflow does not occur outside development
environments, so replace overflow checking with one spot check
at the end. This reduces the number of conditional branches in
the common case.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>

If a request is on the commit list, but is locked, we will currently skip
it, which can lead to livelocking when the commit count doesn't reduce
to zero.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Now that we no longer hold the inode->i_lock when manipulating the
commit lists, it is safe to call pnfs_put_lseg() again.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Switch from using the inode->i_lock for this to avoid contention with
other metadata manipulation.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Rather than forcing us to take the inode->i_lock just in order to bump
the number.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

The commit lists can get very large, so using the inode->i_lock can
end up affecting general metadata performance.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Split out the 2 cases so that we can treat the locking differently.
The issue is that the locking in the pageswapcache cache is highly
linked to the commit list locking.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Hide the locking from nfs_lock_and_join_requests() so that we can
separate out the requirements for swapcache pages.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

Fix up the test in nfs_page_group_covers_page(). The simplest implementation
is to check that we have a set of intersecting or contiguous subrequests
that connect page offset 0 to nfs_page_length(req->wb_page).
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>

nfs_page_group_lock() is now always called with the 'nonblock'
parameter set to 'false'.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>