xprtrdma needs to allocate the Call and Reply buffers separately.
TBH, the reliance on using a single buffer for the pair of XDR
buffers is transport implementation-specific.
Instead of passing just the rq_buffer into the buf_free method, pass
the task structure and let buf_free take care of freeing both
XDR buffers at once.
There's a micro-optimization here. In the common case, both
xprt_release and the transport's buf_free method were checking if
rq_buffer was NULL. Now the check is done only once per RPC.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
xprtrdma needs to allocate the Call and Reply buffers separately.
TBH, the reliance on using a single buffer for the pair of XDR
buffers is transport implementation-specific.
Transports that want to allocate separate Call and Reply buffers
will ignore the "size" argument anyway. Don't bother passing it.
The buf_alloc method can't return two pointers. Instead, make the
method's return value an error code, and set the rq_buffer pointer
in the method itself.
This gives call_allocate an opportunity to terminate an RPC instead
of looping forever when a permanent problem occurs. If a request is
just bogus, or the transport is in a state where it can't allocate
resources for any request, there needs to be a way to kill the RPC
right there and not loop.
This immediately fixes a rare problem in the backchannel send path,
which loops if the server happens to send a CB request whose
call+reply size is larger than a page (which it shouldn't do yet).
One more issue: looks like xprt_inject_disconnect was incorrectly
placed in the failure path in call_allocate. It needs to be in the
success path, as it is for other call-sites.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Use the low latency transport workqueue to process the task that is
next in line on the xprt->sending queue.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
rpciod can easily get congested due to the long list of queued rpc_tasks.
Having the receive queue wait in turn for those tasks to complete can
therefore be a bottleneck.
Address the problem by separating the workqueues into:
- rpciod: manages rpc_tasks
- xprtiod: manages transport related work.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
This is needed in order to allow the NFSv4.1 backchannel and
BIND_CONN_TO_SESSION function to work.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Jerome reported seeing a warning pop when working with a swapfile on
NFS. The nfs_swap_activate can end up calling sk_set_memalloc while
holding the rcu_read_lock and that function can sleep.
To fix that, we need to take a reference to the xprt while holding the
rcu_read_lock, set the socket up for swapping and then drop that
reference. But, xprt_put is not exported and having NFS deal with the
underlying xprt is a bit of layering violation anyway.
Fix this by adding a set of activate/deactivate functions that take a
rpc_clnt pointer instead of an rpc_xprt, and have nfs_swap_activate and
nfs_swap_deactivate call those.
Also, add a per-rpc_clnt atomic counter to keep track of the number of
active swapfiles associated with it. When the counter does a 0->1
transition, we enable swapping on the xprt, when we do a 1->0 transition
we disable swapping on it.
This also allows us to be a bit more selective with the RPC_TASK_SWAPPER
flag. If non-swapper and swapper clnts are sharing a xprt, then we only
need to flag the tasks from the swapper clnt with that flag.
Acked-by: Mel Gorman <mgorman@suse.de>
Reported-by: Jerome Marchand <jmarchan@redhat.com>
Signed-off-by: Jeff Layton <jeff.layton@primarydata.com>
Reviewed-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
It's always set to the same value as CONFIG_TRACEPOINTS, so we can just
use that instead.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
It's always set to whatever CONFIG_SUNRPC_DEBUG is, so just use that.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
It is currently not possible for various wait_on_bit functions
to implement a timeout.
While the "action" function that is called to do the waiting
could certainly use schedule_timeout(), there is no way to carry
forward the remaining timeout after a false wake-up.
As false-wakeups a clearly possible at least due to possible
hash collisions in bit_waitqueue(), this is a real problem.
The 'action' function is currently passed a pointer to the word
containing the bit being waited on. No current action functions
use this pointer. So changing it to something else will be a
little noisy but will have no immediate effect.
This patch changes the 'action' function to take a pointer to
the "struct wait_bit_key", which contains a pointer to the word
containing the bit so nothing is really lost.
It also adds a 'private' field to "struct wait_bit_key", which
is initialized to zero.
An action function can now implement a timeout with something
like
static int timed_out_waiter(struct wait_bit_key *key)
{
unsigned long waited;
if (key->private == 0) {
key->private = jiffies;
if (key->private == 0)
key->private -= 1;
}
waited = jiffies - key->private;
if (waited > 10 * HZ)
return -EAGAIN;
schedule_timeout(waited - 10 * HZ);
return 0;
}
If any other need for context in a waiter were found it would be
easy to use ->private for some other purpose, or even extend
"struct wait_bit_key".
My particular need is to support timeouts in nfs_release_page()
to avoid deadlocks with loopback mounted NFS.
While wait_on_bit_timeout() would be a cleaner interface, it
will not meet my need. I need the timeout to be sensitive to
the state of the connection with the server, which could change.
So I need to use an 'action' interface.
Signed-off-by: NeilBrown <neilb@suse.de>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Steve French <sfrench@samba.org>
Cc: David Howells <dhowells@redhat.com>
Cc: Steven Whitehouse <swhiteho@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/20140707051604.28027.41257.stgit@notabene.brown
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Highlights include:
- Fix NFSv4 recovery so that it doesn't recover lost locks in cases such as
lease loss due to a network partition, where doing so may result in data
corruption. Add a kernel parameter to control choice of legacy behaviour
or not.
- Performance improvements when 2 processes are writing to the same file.
- Flush data to disk when an RPCSEC_GSS session timeout is imminent.
- Implement NFSv4.1 SP4_MACH_CRED state protection to prevent other
NFS clients from being able to manipulate our lease and file lockingr
state.
- Allow sharing of RPCSEC_GSS caches between different rpc clients
- Fix the broken NFSv4 security auto-negotiation between client and server
- Fix rmdir() to wait for outstanding sillyrename unlinks to complete
- Add a tracepoint framework for debugging NFSv4 state recovery issues.
- Add tracing to the generic NFS layer.
- Add tracing for the SUNRPC socket connection state.
- Clean up the rpc_pipefs mount/umount event management.
- Merge more patches from Chuck in preparation for NFSv4 migration support.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.14 (GNU/Linux)
iQIcBAABAgAGBQJSLelVAAoJEGcL54qWCgDyo2IQAKOfRJyZVnf4ipxi3xLNl1QF
w/70DVSIF1S1djWN7G3vgkxj/R8KCvJ8CcvkAD2BEgRDeZJ9TtyKAdM/jYLZ+W05
7k2QKk8fkwZmc1Y2qDqFwKHzP5ZgP5L2nGx7FNhi/99wEAe47yFG3qd3rUWKrcOf
mnd863zgGDE2Q10slhoq/bywwMJo6tKZNeaIE8kPjgFbBEh/jslpAWr8dSA4QgvJ
nZ8VB5XU8L+XJ0GpHHdjYm9LvQ51DbQ6omOF+0P4fI093azKmf4ZsrjMDWT8+iu3
XkXlnQmKLGTi7yB43hHtn2NiRqwGzCcZ1Amo9PpCFaHUt1RP9cc37UhG1T+x1xWJ
STEKDbvCdQ3FU9FvbgrGEwBR0e8fNS4fZY3ToDBflIcfwre0aWs5RCodZMUD0nUI
4wY5J9NsQR/bL+v8KeUR4V4cXK8YrgL0zB4u4WYzH5Npxr5KD0NEKDNqRPhrB9l2
LLF9Haql8j76Ff0ek6UGFIZjDE0h6Fs71wLBpLj+ZWArOJ7vBuLMBSOVqNpld9+9
f2fEG7qoGF4FGTY4myH/eakMPaWnk9Ol4Ls/svSIapJ9+rePD+a93e/qnmdofIMf
4TuEYk6ERib1qXgaeDRQuCsm2YE1Co5skGMaOsRFWgReE1c12QoJQVst2nMtEKp3
uV2w8LgX18aZOZXJVkCM
=ZuW+
-----END PGP SIGNATURE-----
Merge tag 'nfs-for-3.12-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs
Pull NFS client updates from Trond Myklebust:
"Highlights include:
- Fix NFSv4 recovery so that it doesn't recover lost locks in cases
such as lease loss due to a network partition, where doing so may
result in data corruption. Add a kernel parameter to control
choice of legacy behaviour or not.
- Performance improvements when 2 processes are writing to the same
file.
- Flush data to disk when an RPCSEC_GSS session timeout is imminent.
- Implement NFSv4.1 SP4_MACH_CRED state protection to prevent other
NFS clients from being able to manipulate our lease and file
locking state.
- Allow sharing of RPCSEC_GSS caches between different rpc clients.
- Fix the broken NFSv4 security auto-negotiation between client and
server.
- Fix rmdir() to wait for outstanding sillyrename unlinks to complete
- Add a tracepoint framework for debugging NFSv4 state recovery
issues.
- Add tracing to the generic NFS layer.
- Add tracing for the SUNRPC socket connection state.
- Clean up the rpc_pipefs mount/umount event management.
- Merge more patches from Chuck in preparation for NFSv4 migration
support"
* tag 'nfs-for-3.12-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs: (107 commits)
NFSv4: use mach cred for SECINFO_NO_NAME w/ integrity
NFS: nfs_compare_super shouldn't check the auth flavour unless 'sec=' was set
NFSv4: Allow security autonegotiation for submounts
NFSv4: Disallow security negotiation for lookups when 'sec=' is specified
NFSv4: Fix security auto-negotiation
NFS: Clean up nfs_parse_security_flavors()
NFS: Clean up the auth flavour array mess
NFSv4.1 Use MDS auth flavor for data server connection
NFS: Don't check lock owner compatability unless file is locked (part 2)
NFS: Don't check lock owner compatibility in writes unless file is locked
nfs4: Map NFS4ERR_WRONG_CRED to EPERM
nfs4.1: Add SP4_MACH_CRED write and commit support
nfs4.1: Add SP4_MACH_CRED stateid support
nfs4.1: Add SP4_MACH_CRED secinfo support
nfs4.1: Add SP4_MACH_CRED cleanup support
nfs4.1: Add state protection handler
nfs4.1: Minimal SP4_MACH_CRED implementation
SUNRPC: Replace pointer values with task->tk_pid and rpc_clnt->cl_clid
SUNRPC: Add an identifier for struct rpc_clnt
SUNRPC: Ensure rpc_task->tk_pid is available for tracepoints
...
If rpcbind causes our connection to the AF_LOCAL socket to close after
we've registered a service, then we want to be careful about reconnecting
since the mount namespace may have changed.
By simply refusing to reconnect the AF_LOCAL socket in the case of
unregister, we avoid the need to somehow save the mount namespace. While
this may lead to some services not unregistering properly, it should
be safe.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: Nix <nix@esperi.org.uk>
Cc: Jeff Layton <jlayton@redhat.com>
Cc: stable@vger.kernel.org # 3.9.x
Currently, the priority queues attempt to be 'fair' to lower priority
tasks by scheduling them after a certain number of higher priority tasks
have run. The problem is that both the transport send queue and
the NFSv4.1 session slot queue have strong ordering requirements.
This patch therefore removes the fairness code in favour of strong
ordering of task priorities.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Include RPC statistics from all data servers in /proc/self/mountstats for pNFS
filelayout mounts.
Signed-off-by: Weston Andros Adamson <dros@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
The tracepoint code relies on the queue->name being defined in order to
be able to display the name of the waitqueue on which an RPC task is
sleeping.
Reported-by: Randy Dunlap <rdunlap@xenotime.net>
Reported-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
Acked-by: Randy Dunlap <rdunlap@xenotime.net>
We have to ensure that the wake up from the waitqueue and the assignment
of xprt->snd_task are atomic. We can do this by assigning the snd_task
while under the waitqueue spinlock.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
This patch moves static SUNRPC clients list and it's lock to sunrpc_net
structure.
Currently this list is used only for debug purposes. But later it will be used
also for selecting clients by networks namespace on PipeFS mount/umount events.
Per-network namespace lists will make this faster and simplier.
Note: client list is taken from "init_net" network namespace context in
rpc_show_tasks(). This will be changed some day later with making SUNRPC
sysctl's per network namespace context.
Signed-off-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Currently, the caller has to change the value of task->tk_priority if
it wants to select on which priority level the task will sleep.
This patch allows the caller to select a priority level at sleep time
rather than always using task->tk_priority.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
If the NLM daemon is killed on the NFS server, we can currently end up
hanging forever on an 'unlock' request, instead of aborting. Basically,
if the rpcbind request fails, or the server keeps returning garbage, we
really want to quit instead of retrying.
Tested-by: Vasily Averin <vvs@sw.ru>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: stable@kernel.org
On occasion, it is useful for the NFS layer to distinguish between
soft timeouts and other EIO errors due to (say) encoding errors,
or authentication errors.
The following patch ensures that the default behaviour of the RPC
layer remains to return EIO on soft timeouts (until we have
audited all the callers).
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
If we fail to contact the gss upcall program, then no message will
be sent to the server. The client still updated the sequence number,
however, and this lead to NFS4ERR_SEQ_MISMATCH for the next several
RPC calls.
Signed-off-by: Bryan Schumaker <bjschuma@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Although they run as rpciod background tasks, under normal operation
(i.e. no SIGKILL), functions like nfs_sillyrename(), nfs4_proc_unlck()
and nfs4_do_close() want to be fully synchronous. This means that when we
exit, we want all references to the rpc_task to be gone, and we want
any dentry references etc. held by that task to be released.
For this reason these functions call __rpc_wait_for_completion_task(),
followed by rpc_put_task() in the expectation that the latter will be
releasing the last reference to the rpc_task, and thus ensuring that the
callback_ops->rpc_release() has been called synchronously.
This patch fixes a race which exists due to the fact that
rpciod calls rpc_complete_task() (in order to wake up the callers of
__rpc_wait_for_completion_task()) and then subsequently calls
rpc_put_task() without ensuring that these two steps are done atomically.
In order to avoid adding new spin locks, the patch uses the existing
waitqueue spin lock to order the rpc_task reference count releases between
the waiting process and rpciod.
The common case where nobody is waiting for completion is optimised for by
checking if the RPC_TASK_ASYNC flag is cleared and/or if the rpc_task
reference count is 1: in those cases we drop trying to grab the spin lock,
and immediately free up the rpc_task.
Those few processes that need to put the rpc_task from inside an
asynchronous context and that do not care about ordering are given a new
helper: rpc_put_task_async().
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Make rpc_exit() non-inline, and ensure that it always wakes up a task that
has been queued.
Kill off the now unused rpc_wake_up_task().
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
This improves the packing of the rpc_task, and ensures that on 64-bit
platforms the size reduces to 216 bytes.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
It seems strange to maintain stats for bytes_sent in one structure, and
bytes received in another. Try to assemble all the RPC request-related
stats in struct rpc_rqst
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Currently RPC performance metrics that tabulate elapsed time use
jiffies time values. This is problematic on systems that use slow
jiffies (for instance 100HZ systems built for paravirtualized
environments). It is also a problem for computing precise latency
statistics for advanced network transports, such as InfiniBand,
that can have round-trip latencies significanly faster than a single
clock tick.
For the RPC client, adopt the high resolution time stamp mechanism
already used by the network layer and blktrace: ktime.
We use ktime format time stamps for all internal computations, and
convert to milliseconds for presentation. As a result, we need only
addition operations in the performance critical paths; multiply/divide
is required only for presentation.
We could report RTT metrics in microseconds. In fact the mountstats
format is versioned to accomodate exactly this kind of interface
improvement.
For now, however, we'll stay with millisecond precision for
presentation to maintain backwards compatibility with the handful of
currently deployed user space tools. At a later point, we'll move to
an API such as BDI_STATS where a finer timestamp precision can be
reported.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
The kernel sometimes makes RPC calls to services that aren't running.
Because the kernel's RPC client always assumes the hard retry semantic
when reconnecting a connection-oriented RPC transport, the underlying
reconnect logic takes a long while to time out, even though the remote
may have responded immediately with ECONNREFUSED.
In certain cases, like upcalls to our local rpcbind daemon, or for NFS
mount requests, we'd like the kernel to fail immediately if the remote
service isn't reachable. This allows another transport to be tried
immediately, or the pending request can be abandoned quickly.
Introduce a per-request flag which controls how call_transmit_status()
behaves when request transmission fails because the server cannot be
reached.
We don't want soft connection semantics to apply to other errors. The
default case of the switch statement in call_transmit_status() no
longer falls through; the fall through code is copied to the default
case, and a "break;" is added.
The transport's connection re-establishment timeout is also ignored for
such requests. We want the request to fail immediately, so the
reconnect delay is skipped. Additionally, we don't want a connect
failure here to further increase the reconnect timeout value, since
this request will not be retried.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Adds rpc_run_bc_task() which is called by the NFS callback service to
process backchannel requests. It performs similar work to rpc_run_task()
though "schedules" the backchannel task to be executed starting at the
call_trasmit state in the RPC state machine.
It also introduces some miscellaneous updates to the argument validation,
call_transmit, and transport cleanup functions to take into account
that there are now forechannel and backchannel tasks.
Backchannel requests do not carry an RPC message structure, since the
payload has already been XDR encoded using the existing NFSv4 callback
mechanism.
Introduce a new transmit state for the client to reply on to backchannel
requests. This new state simply reserves the transport and issues the
reply. In case of a connection related error, disconnects the transport and
drops the reply. It requires the forechannel to re-establish the connection
and the server to retransmit the request, as stated in NFSv4.1 section
2.9.2 "Client and Server Transport Behavior".
Note: There is no need to loop attempting to reserve the transport. If EAGAIN
is returned by xprt_prepare_transmit(), return with tk_status == 0,
setting tk_action to call_bc_transmit. rpc_execute() will invoke it again
after the task is taken off the sleep queue.
[nfs41: rpc_run_bc_task() need not be exported outside RPC module]
[nfs41: New call_bc_transmit RPC state]
Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com>
Signed-off-by: Benny Halevy <bhalevy@panasas.com>
[nfs41: Backchannel: No need to loop in call_bc_transmit()]
Signed-off-by: Andy Adamson <andros@netapp.com>
Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com>
Signed-off-by: Benny Halevy <bhalevy@panasas.com>
[rpc_count_iostats incorrectly exits early]
Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com>
Signed-off-by: Benny Halevy <bhalevy@panasas.com>
[Convert rpc_reply_expected() to inline function]
[Remove unnecessary BUG_ON()]
[Rename variable]
Signed-off-by: Ricardo Labiaga <Ricardo.Labiaga@netapp.com>
Signed-off-by: Benny Halevy <bhalevy@panasas.com>
Try to make the comment here a little more clear and concise.
Also, this macro definition seems unnecessary.
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Now that we've tightened up the locking rules for RPC queue wakeups, we can
remove the RCU-safe kfree calls...
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
This is designed to replace the timeout timer in the individual rpc_tasks.
By putting the timer function in the wait queue, we will eventually be able
to reduce the total number of timers in use by the RPC subsystem.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
An audit of the current RPC timeout functions shows that they don't really
ever need to run in the softirq context. As long as the softirq is
able to signal that the wakeup is due to a timeout (which it can do by
setting task->tk_status to -ETIMEDOUT) then the callback functions can just
run as standard task->tk_callback functions (in the rpciod/process
context).
The only possible border-line case would be xprt_timer() for the case of
UDP, when the callback is used to reduce the size of the transport
congestion window. In testing, however, the effect of moving that update
to a callback would appear to be minor.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Chuck Lever