An RPC/RDMA client may send large RPC arguments via a read
list. This is a list of scatter/gather elements which convey
RPC call arguments too large to fit in a small RDMA SEND.
Each entry in the read list has a "position" field, whose value is
the byte offset in the XDR stream where the data in that entry is to
be inserted. Entries which share the same "position" value make up
the same RPC argument. The receiver inserts entries with the same
position field value in list order into the XDR stream.
Currently the Linux NFS/RDMA server cannot handle receiving read
chunks in more than one position, mostly because no current client
sends read lists with elements in more than one position. As a
sanity check, ensure that all received chunks have the same
"rc_position."
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
The RDMA reader function doesn't change once an svcxprt_rdma is
instantiated. Instead of checking sc_devcap during every incoming
RPC, set the reader function once when the connection is accepted.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
xdr_start() can return the wrong rmsgp address if an assumption
about how the xdr_buf was constructed changes. When it gets it
wrong, the client receives a reply that has gibberish in the
RPC/RDMA header, preventing it from matching a waiting RPC request.
Instead, make (and document) just one assumption: that the RDMA
header for the client's RPC call is at the start of the first page
in rq_pages.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Current convention is to avoid using BUG_ON() in places where an
oops could cause complete system failure.
Replace BUG_ON() call sites in svcrdma with an assertion error
message and allow execution to continue safely.
Some BUG_ON() calls are removed because they have never fired in
production (that we are aware of).
Some WARN_ON() calls are also replaced where a back trace is not
helpful; e.g., in a workqueue task.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
The byte_count argument is not used, and the function is called
only from one place.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Nit: remove an unused variable to squelch a compiler warning.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Nit: Fix inconsistent white space in dprintk messages.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
A struct xdr_stream at a page boundary might point to the end of one
page or the beginning of the next, but xdr_truncate_encode isn't
prepared to handle the former.
This can cause corruption of NFSv4 READDIR replies in the case that a
readdir entry that would have exceeded the client's dircount/maxcount
limit would have ended exactly on a 4k page boundary. You're more
likely to hit this case on large directories.
Other xdr_truncate_encode callers are probably also affected.
Reported-by: Holger Hoffstätte <holger.hoffstaette@googlemail.com>
Tested-by: Holger Hoffstätte <holger.hoffstaette@googlemail.com>
Fixes: 3e19ce762b "rpc: xdr_truncate_encode"
Cc: stable@vger.kernel.org
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
There is nice kernel helper to escape a given strings by provided rules. Let's
use it instead of custom approach.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
[bfields@redhat.com: fix length calculation]
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
...move the WARN_ON_ONCE inside the following if block since they use
the same condition.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
These were useful when I was tracking down a race condition between
svc_xprt_do_enqueue and svc_get_next_xprt.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Testing has shown that the pool->sp_lock can be a bottleneck on a busy
server. Every time data is received on a socket, the server must take
that lock in order to dequeue a thread from the sp_threads list.
Address this problem by eliminating the sp_threads list (which contains
threads that are currently idle) and replacing it with a RQ_BUSY flag in
svc_rqst. This allows us to walk the sp_all_threads list under the
rcu_read_lock and find a suitable thread for the xprt by doing a
test_and_set_bit.
Note that we do still have a potential atomicity problem however with
this approach. We don't want svc_xprt_do_enqueue to set the
rqst->rq_xprt pointer unless a test_and_set_bit of RQ_BUSY returned
zero (which indicates that the thread was idle). But, by the time we
check that, the bit could be flipped by a waking thread.
To address this, we acquire a new per-rqst spinlock (rq_lock) and take
that before doing the test_and_set_bit. If that returns false, then we
can set rq_xprt and drop the spinlock. Then, when the thread wakes up,
it must set the bit under the same spinlock and can trust that if it was
already set then the rq_xprt is also properly set.
With this scheme, the case where we have an idle thread no longer needs
to take the highly contended pool->sp_lock at all, and that removes the
bottleneck.
That still leaves one issue: What of the case where we walk the whole
sp_all_threads list and don't find an idle thread? Because the search is
lockess, it's possible for the queueing to race with a thread that is
going to sleep. To address that, we queue the xprt and then search again.
If we find an idle thread at that point, we can't attach the xprt to it
directly since that might race with a different thread waking up and
finding it. All we can do is wake the idle thread back up and let it
attempt to find the now-queued xprt.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Tested-by: Chris Worley <chris.worley@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
In a later patch, we'll be removing some spinlocking around the socket
and thread queueing code in order to fix some contention problems. At
that point, the stats counters will no longer be protected by the
sp_lock.
Change the counters to atomic_long_t fields, except for the
"sockets_queued" counter which will still be manipulated under a
spinlock.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Tested-by: Chris Worley <chris.worley@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
...also make the manipulation of sp_all_threads list use RCU-friendly
functions.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Tested-by: Chris Worley <chris.worley@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Currently all svc_create callers pass in NULL for the shutdown parm,
which then gets fixed up to be svc_rpcb_cleanup if the service uses
rpcbind.
Simplify this by instead having the the only caller that requires it
(lockd) pass in svc_rpcb_cleanup and get rid of the special casing.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
The way that svc_wake_up works is a bit inefficient. It walks all of the
available pools for a service and either wakes up a task in each one or
sets the SP_TASK_PENDING flag in each one.
When svc_wake_up is called, there is no need to wake up more than one
thread to do this work. In practice, only lockd currently uses this
function and it's single threaded anyway. Thus, this just boils down to
doing a wake up of a thread in pool 0 or setting a single flag.
Eliminate the for loop in this function and change it to just operate on
pool 0. Also update the comments that sit above it and get rid of some
code that has been commented out for years now.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
In a later patch, we'll want to be able to handle this flag without
holding the sp_lock. Change this field to an unsigned long flags
field, and declare a new flag in it that can be managed with atomic
bitops.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
In a later patch, we're going to need some atomic bit flags. Since that
field will need to be an unsigned long, we mitigate that space
consumption by migrating some other bitflags to the new field. Start
with the rq_secure flag.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Highlights include:
Features:
- NFSv4.2 client support for hole punching and preallocation.
- Further RPC/RDMA client improvements.
- Add more RPC transport debugging tracepoints.
- Add RPC debugging tools in debugfs.
Bugfixes:
- Stable fix for layoutget error handling
- Fix a change in COMMIT behaviour resulting from the recent io code updates
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=wOP3
-----END PGP SIGNATURE-----
Merge tag 'nfs-for-3.19-1' into nfsd for-3.19 branch
Mainly what I need is 860a0d9e51 "sunrpc: add some tracepoints in
svc_rqst handling functions", which subsequent server rpc patches from
jlayton depend on. I'm merging this later tag on the assumption that's
more likely to be a tested and stable point.
Currently, it leaks when the allocation fails.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
All it does is indicate whether a xprt has already been deleted from
a list or not, which is unnecessary since we use list_del_init and it's
always set and checked under the sv_lock anyway.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Add a new directory heirarchy under the debugfs sunrpc/ directory:
sunrpc/
rpc_xprt/
<xprt id>/
Within that directory, we can put files that give info about the
xprts. We do have the (minor) problem that there is no succinct,
unique identifier for rpc_xprts. So we generate them synthetically
with a static atomic_t counter.
For now, this directory just holds an "info" file, but we may add
other files to it in the future.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
It's possible to get a dump of the RPC task queue by writing a value to
/proc/sys/sunrpc/rpc_debug. If you write any value to that file, you get
a dump of the RPC client task list into the log buffer. This is a rather
inconvenient interface however, and makes it hard to get immediate info
about the task queue.
Add a new directory hierarchy under debugfs:
sunrpc/
rpc_clnt/
<clientid>/
Within each clientid directory we create a new "tasks" file that will
dump info similar to what shows up in the log buffer, but with a few
small differences -- we avoid printing raw kernel addresses in favor of
symbolic names and the XID is also displayed.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
These patches various bugfixes and cleanups for using NFS over RDMA, including
better error handling and performance improvements by using pad optimization.
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
iQIcBAABCAAGBQJUdPv2AAoJENfLVL+wpUDrJkMQAKjtPZHLMcj+eHm4f1ZKLJxy
GSrUZV21TU9tL0NVE/5An8US6hoLwHpNXsW8o+gHTAeGRyiCmIaNXGd1Ql/4PYRH
zfzdNXoaJAh1N5iXX11fF3gOWqx/SolqzO2xLDVETK/3lAvq0VwMYoMElBQB6qQW
8sN3z8yVuz/9Ia9oGIFhqu1B6dcKPHkQDMtmsElGxeEX/+9yEg4HUKx+kZDtV0Uj
8/JM8Jh1FKRCQT/P6INkRItdY5KaSJGFc43BkC/8lbugfxa5XCyu/m/qMr9FJsDV
nM6rwaiVcmR/mvD3fL82+Jg/M+P9VUHQ1/Az0sV9G+fEoHH/1Mey3LfMzNpUmf9v
bykrPRuzXkPPQgN1VnjSaF2RF+CWwV9Nme1VVXM/zj8gHX1mcmQF/wPRxDuLjCrt
EObAFsvHOwDTZZmYp9bG5kc6IvwvT8aeeVQMJ4q4PSGD3w8AtoIyJDn+Ee0LFD1K
Zw0oZpTJpI4t7DVxGBSdo2wZWuMU/UKqGqGtGJ+ljXfTRuuq968Q5j5ujaA9vf0v
C9igYTU8hq4teMzhZrfR1jtTWoSS+5zamb1KtvAZy8gsht2PQVgE9xka2k/AV8uE
ul/w5HU4OV+QIrHNbiu7BE8B2Ags6smpdHMqn9fqLBwvG+JEwbWqk1zeTsajxzq+
hkvKkkMq6JjDbsDf96Yk
=YIru
-----END PGP SIGNATURE-----
Merge tag 'nfs-rdma-for-3.19' of git://git.linux-nfs.org/projects/anna/nfs-rdma into linux-next
Pull NFS client RDMA changes for 3.19 from Anna Schumaker:
"NFS: Client side changes for RDMA
These patches various bugfixes and cleanups for using NFS over RDMA, including
better error handling and performance improvements by using pad optimization.
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>"
* tag 'nfs-rdma-for-3.19' of git://git.linux-nfs.org/projects/anna/nfs-rdma:
xprtrdma: Display async errors
xprtrdma: Enable pad optimization
xprtrdma: Re-write rpcrdma_flush_cqs()
xprtrdma: Refactor tasklet scheduling
xprtrdma: unmap all FMRs during transport disconnect
xprtrdma: Cap req_cqinit
xprtrdma: Return an errno from rpcrdma_register_external()
These patches fixes for iostats and SETCLIENTID in addition to cleaning
up the nfs4_init_callback() function.
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2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=PMdS
-----END PGP SIGNATURE-----
Merge tag 'nfs-cel-for-3.19' of git://git.linux-nfs.org/projects/anna/nfs-rdma into linux-next
Pull pull additional NFS client changes for 3.19 from Anna Schumaker:
"NFS: Generic client side changes from Chuck
These patches fixes for iostats and SETCLIENTID in addition to cleaning
up the nfs4_init_callback() function.
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>"
* tag 'nfs-cel-for-3.19' of git://git.linux-nfs.org/projects/anna/nfs-rdma:
NFS: Clean up nfs4_init_callback()
NFS: SETCLIENTID XDR buffer sizes are incorrect
SUNRPC: serialize iostats updates
Occasionally mountstats reports a negative retransmission rate.
Ensure that two RPCs completing concurrently don't confuse the sums
in the transport's op_metrics array.
Since pNFS filelayout can invoke rpc_count_iostats() on another
transport from xprt_release(), we can't rely on simply holding the
transport_lock in xprt_release(). There's nothing for it but hard
serialization. One spin lock per RPC operation should make this as
painless as it can be.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
An async error upcall is a hard error, and should be reported in
the system log.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
The Linux NFS/RDMA server used to reject NFSv3 WRITE requests when
pad optimization was enabled. That bug was fixed by commit
e560e3b510 ("svcrdma: Add zero padding if the client doesn't send
it").
We can now enable pad optimization on the client, which helps
performance and is supported now by both Linux and Solaris servers.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Currently rpcrdma_flush_cqs() attempts to avoid code duplication,
and simply invokes rpcrdma_recvcq_upcall and rpcrdma_sendcq_upcall.
1. rpcrdma_flush_cqs() can run concurrently with provider upcalls.
Both flush_cqs() and the upcalls were invoking ib_poll_cq() in
different threads using the same wc buffers (ep->rep_recv_wcs
and ep->rep_send_wcs), added by commit 1c00dd0776 ("xprtrmda:
Reduce calls to ib_poll_cq() in completion handlers").
During transport disconnect processing, this sometimes resulted
in the same reply getting added to the rpcrdma_tasklets_g list
more than once, which corrupted the list.
2. The upcall functions drain only a limited number of CQEs,
thanks to the poll budget added by commit 8301a2c047
("xprtrdma: Limit work done by completion handler").
Fixes: a7bc211ac9 ("xprtrdma: On disconnect, don't ignore ... ")
BugLink: https://bugzilla.linux-nfs.org/show_bug.cgi?id=276
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Restore the separate function that schedules the reply handling
tasklet. I need to call it from two different paths.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
When using RPCRDMA_MTHCAFMR memory registration, after a few
transport disconnect / reconnect cycles, ib_map_phys_fmr() starts to
return EINVAL because the provider has exhausted its map pool.
Make sure that all FMRs are unmapped during transport disconnect,
and that ->send_request remarshals them during an RPC retransmit.
This resets the transport's MRs to ensure that none are leaked
during a disconnect.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Recent work made FRMR registration and invalidation completions
unsignaled. This greatly reduces the adapter interrupt rate.
Every so often, however, a posted send Work Request is allowed to
signal. Otherwise, the provider's Work Queue will wrap and the
workload will hang.
The number of Work Requests that are allowed to remain unsignaled is
determined by the value of req_cqinit. Currently, this is set to the
size of the send Work Queue divided by two, minus 1.
For FRMR, the send Work Queue is the maximum number of concurrent
RPCs (currently 32) times the maximum number of Work Requests an
RPC might use (currently 7, though some adapters may need more).
For mlx4, this is 224 entries. This leaves completion signaling
disabled for 111 send Work Requests.
Some providers hold back dispatching Work Requests until a CQE is
generated. If completions are disabled, then no CQEs are generated
for quite some time, and that can stall the Work Queue.
I've seen this occur running xfstests generic/113 over NFSv4, where
eventually, posting a FAST_REG_MR Work Request fails with -ENOMEM
because the Work Queue has overflowed. The connection is dropped
and re-established.
Cap the rep_cqinit setting so completions are not left turned off
for too long.
BugLink: https://bugzilla.linux-nfs.org/show_bug.cgi?id=269
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
The RPC/RDMA send_request method and the chunk registration code
expects an errno from the registration function. This allows
the upper layers to distinguish between a recoverable failure
(for example, temporary memory exhaustion) and a hard failure
(for example, a bug in the registration logic).
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.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>
Add tracepoints inside the main loop on xs_tcp_data_recv that allow
us to keep an eye on what's happening during each phase of it.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
...so we can keep track of when calls are sent and replies received.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
...just around svc_send, svc_recv and svc_process for now.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Both xprt_lookup_rqst() and xprt_complete_rqst() require that you
take the transport lock in order to avoid races with xprt_transmit().
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Cc: stable@vger.kernel.org
Reviewed-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
The m->pool_to[] array has "maxpools" number of elements. It's
allocated in svc_pool_map_alloc_arrays() which we called earlier in the
function. This test should be >= instead of >.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Thanks to Andrea Arcangeli for pointing out these checks are
obviously unnecessary given the preceding calculations.
Reported-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Pull nfsd updates from Bruce Fields:
"Highlights:
- support the NFSv4.2 SEEK operation (allowing clients to support
SEEK_HOLE/SEEK_DATA), thanks to Anna.
- end the grace period early in a number of cases, mitigating a
long-standing annoyance, thanks to Jeff
- improve SMP scalability, thanks to Trond"
* 'for-3.18' of git://linux-nfs.org/~bfields/linux: (55 commits)
nfsd: eliminate "to_delegation" define
NFSD: Implement SEEK
NFSD: Add generic v4.2 infrastructure
svcrdma: advertise the correct max payload
nfsd: introduce nfsd4_callback_ops
nfsd: split nfsd4_callback initialization and use
nfsd: introduce a generic nfsd4_cb
nfsd: remove nfsd4_callback.cb_op
nfsd: do not clear rpc_resp in nfsd4_cb_done_sequence
nfsd: fix nfsd4_cb_recall_done error handling
nfsd4: clarify how grace period ends
nfsd4: stop grace_time update at end of grace period
nfsd: skip subsequent UMH "create" operations after the first one for v4.0 clients
nfsd: set and test NFSD4_CLIENT_STABLE bit to reduce nfsdcltrack upcalls
nfsd: serialize nfsdcltrack upcalls for a particular client
nfsd: pass extra info in env vars to upcalls to allow for early grace period end
nfsd: add a v4_end_grace file to /proc/fs/nfsd
lockd: add a /proc/fs/lockd/nlm_end_grace file
nfsd: reject reclaim request when client has already sent RECLAIM_COMPLETE
nfsd: remove redundant boot_time parm from grace_done client tracking op
...
Svcrdma currently advertises 1MB, which is too large. The correct value
is the minimum of RPCSVC_MAXPAYLOAD and the max scatter-gather allowed
in an NFSRDMA IO chunk * the host page size. This bug is usually benign
because the Linux X64 NFSRDMA client correctly limits the payload size to
the correct value (64*4096 = 256KB). But if the Linux client is PPC64
with a 64KB page size, then the client will indeed use a payload size
that will overflow the server.
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
The flag RPC_CLNT_CREATE_NO_RETRANS_TIMEOUT was intended introduced in
order to allow NFSv4 clients to disable resend timeouts. Since those
cause the RPC layer to break the connection, they mess up the duplicate
reply caches that remain indexed on the port number in NFSv4..
This patch includes the code that was missing in the original to
set the appropriate flag in struct rpc_clnt, when the caller of
rpc_create() sets RPC_CLNT_CREATE_NO_RETRANS_TIMEOUT.
Fixes: 8a19a0b6cb (SUNRPC: Add RPC task and client level options to...)
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Now that nfs_release_page() doesn't block indefinitely, other deadlock
avoidance mechanisms aren't needed.
- it doesn't hurt for kswapd to block occasionally. If it doesn't
want to block it would clear __GFP_WAIT. The current_is_kswapd()
was only added to avoid deadlocks and we have a new approach for
that.
- memory allocation in the SUNRPC layer can very rarely try to
->releasepage() a page it is trying to handle. The deadlock
is removed as nfs_release_page() doesn't block indefinitely.
So we don't need to set PF_FSTRANS for sunrpc network operations any
more.
Signed-off-by: NeilBrown <neilb@suse.de>
Acked-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
If an iptables drop rule is added for an nfs server, the client can end up in
a softlockup. Because of the way that xs_sendpages() is structured, the -EPERM
is ignored since the prior bits of the packet may have been successfully queued
and thus xs_sendpages() returns a non-zero value. Then, xs_udp_send_request()
thinks that because some bits were queued it should return -EAGAIN. We then try
the request again and again, resulting in cpu spinning. Reproducer:
1) open a file on the nfs server '/nfs/foo' (mounted using udp)
2) iptables -A OUTPUT -d <nfs server ip> -j DROP
3) write to /nfs/foo
4) close /nfs/foo
5) iptables -D OUTPUT -d <nfs server ip> -j DROP
The softlockup occurs in step 4 above.
The previous patch, allows xs_sendpages() to return both a sent count and
any error values that may have occurred. Thus, if we get an -EPERM, return
that to the higher level code.
With this patch in place we can successfully abort the above sequence and
avoid the softlockup.
I also tried the above test case on an nfs mount on tcp and although the system
does not softlockup, I still ended up with the 'hung_task' firing after 120
seconds, due to the i/o being stuck. The tcp case appears a bit harder to fix,
since -EPERM appears to get ignored much lower down in the stack and does not
propogate up to xs_sendpages(). This case is not quite as insidious as the
softlockup and it is not addressed here.
Reported-by: Yigong Lou <ylou@akamai.com>
Signed-off-by: Jason Baron <jbaron@akamai.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
If an error is returned after the first bits of a packet have already been
successfully queued, xs_sendpages() will return a positive 'int' value
indicating success. Callers seem to treat this as -EAGAIN.
However, there are cases where its not a question of waiting for the write
queue to drain. For example, when there is an iptables rule dropping packets
to the destination, the lower level code can return -EPERM only after parts
of the packet have been successfully queued. In this case, we can end up
continuously retrying resulting in a kernel softlockup.
This patch is intended to make no changes in behavior but is in preparation for
subsequent patches that can make decisions based on both on the number of bytes
sent by xs_sendpages() and any errors that may have be returned.
Signed-off-by: Jason Baron <jbaron@akamai.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
When aborting a connection to preserve source ports, don't wake the task in
xs_error_report. This allows tasks with RPC_TASK_SOFTCONN to succeed if the
connection needs to be re-established since it preserves the task's status
instead of setting it to the status of the aborting kernel_connect().
This may also avoid a potential conflict on the socket's lock.
Signed-off-by: Benjamin Coddington <bcodding@redhat.com>
Cc: stable@vger.kernel.org # 3.14+
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
When attempting to establish a local ephemeral endpoint for a TCP or UDP
socket, do not explicitly call bind, instead let it happen implicilty when the
socket is first used.
The main motivating factor for this change is when TCP runs out of unique
ephemeral ports (i.e. cannot find any ephemeral ports which are not a part of
*any* TCP connection). In this situation if you explicitly call bind, then the
call will fail with EADDRINUSE. However, if you allow the allocation of an
ephemeral port to happen implicitly as part of connect (or other functions),
then ephemeral ports can be reused, so long as the combination of (local_ip,
local_port, remote_ip, remote_port) is unique for TCP sockets on the system.
This doesn't matter for UDP sockets, but it seemed easiest to treat TCP and UDP
sockets the same.
This can allow mount.nfs(8) to continue to function successfully, even in the
face of misbehaving applications which are creating a large number of TCP
connections.
Signed-off-by: Chris Perl <chris.perl@gmail.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
xprt_lookup_rqst() and bc_send_request() display a byte-swapped XID,
but receive_cb_reply() does not.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
current_task appears to be x86-only, oops.
Let's just delete this check entirely:
Any developer that adds a new user without setting rq_task will get a
crash the first time they test it. I also don't think there are
normally any important locks held here, and I can't see any other reason
why killing a server thread would bring the whole box down.
So the effort to fail gracefully here looks like overkill.
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Fixes: 983c684466 "SUNRPC: get rid of the request wait queue"
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
We really do not want to do ioctls in the server's fast path. Instead, let's
use the fact that we managed to read a full record as the indicator that
we should try to read the socket again.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Just move the transport locking out of the spin lock protected area
altogether.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
We should definitely not be exiting svc_get_next_xprt() with the
thread enqueued. Fix this by ensuring that we fall through to
the dequeue.
Also move the test itself outside the spin lock protected section.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
We're always _only_ waking up tasks from within the sp_threads list, so
we know that they are enqueued and alive. The rq_wait waitqueue is just
a distraction with extra atomic semantics.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
We already determined that there was enough wspace when we
called svc_xprt_enqueue.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Highlights include:
- Stable fix for a bug in nfs3_list_one_acl()
- Speed up NFS path walks by supporting LOOKUP_RCU
- More read/write code cleanups
- pNFS fixes for layout return on close
- Fixes for the RCU handling in the rpcsec_gss code
- More NFS/RDMA fixes
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1
iQIcBAABAgAGBQJT65zoAAoJEGcL54qWCgDyvq8QAJ+OKuC5dpngrZ13i4ZJIcK1
TJSkWCr44FhYPlrmkLCntsGX6C0376oFEtJ5uqloqK0+/QtvwRNVSQMKaJopKIVY
mR4En0WwpigxVQdW2lgto6bfOhzMVO+llVdmicEVrU8eeSThATxGNv7rxRzWorvL
RX3TwBkWSc0kLtPi66VRFQ1z+gg5I0kngyyhsKnLOaHHtpTYP2JDZlRPRkokXPUg
nmNedmC3JrFFkarroFIfYr54Qit2GW/eI2zVhOwHGCb45j4b2wntZ6wr7LpUdv3A
OGDBzw59cTpcx3Hij9CFvLYVV9IJJHBNd2MJqdQRtgWFfs+aTkZdk4uilUJCIzZh
f4BujQAlm/4X1HbPxsSvkCRKga7mesGM7e0sBDPHC1vu0mSaY1cakcj2kQLTpbQ7
gqa1cR3pZ+4shCq37cLwWU0w1yElYe1c4otjSCttPCrAjXbXJZSFzYnHm8DwKROR
t+yEDRL5BIXPu1nEtSnD2+xTQ3vUIYXooZWEmqLKgRtBTtPmgSn9Vd8P1OQXmMNo
VJyFXyjNx5WH06Wbc/jLzQ1/cyhuPmJWWyWMJlVROyv+FXk9DJUFBZuTkpMrIPcF
NlBXLV1GnA7PzMD9Xt9bwqteERZl6fOUDJLWS9P74kTk5c2kD+m+GaqC/rBTKKXc
ivr2s7aIDV48jhnwBSVL
=KE07
-----END PGP SIGNATURE-----
Merge tag 'nfs-for-3.17-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs
Pull NFS client updates from Trond Myklebust:
"Highlights include:
- stable fix for a bug in nfs3_list_one_acl()
- speed up NFS path walks by supporting LOOKUP_RCU
- more read/write code cleanups
- pNFS fixes for layout return on close
- fixes for the RCU handling in the rpcsec_gss code
- more NFS/RDMA fixes"
* tag 'nfs-for-3.17-1' of git://git.linux-nfs.org/projects/trondmy/linux-nfs: (79 commits)
nfs: reject changes to resvport and sharecache during remount
NFS: Avoid infinite loop when RELEASE_LOCKOWNER getting expired error
SUNRPC: remove all refcounting of groupinfo from rpcauth_lookupcred
NFS: fix two problems in lookup_revalidate in RCU-walk
NFS: allow lockless access to access_cache
NFS: teach nfs_lookup_verify_inode to handle LOOKUP_RCU
NFS: teach nfs_neg_need_reval to understand LOOKUP_RCU
NFS: support RCU_WALK in nfs_permission()
sunrpc/auth: allow lockless (rcu) lookup of credential cache.
NFS: prepare for RCU-walk support but pushing tests later in code.
NFS: nfs4_lookup_revalidate: only evaluate parent if it will be used.
NFS: add checks for returned value of try_module_get()
nfs: clear_request_commit while holding i_lock
pnfs: add pnfs_put_lseg_async
pnfs: find swapped pages on pnfs commit lists too
nfs: fix comment and add warn_on for PG_INODE_REF
nfs: check wait_on_bit_lock err in page_group_lock
sunrpc: remove "ec" argument from encrypt_v2 operation
sunrpc: clean up sparse endianness warnings in gss_krb5_wrap.c
sunrpc: clean up sparse endianness warnings in gss_krb5_seal.c
...
Pull nfsd updates from Bruce Fields:
"This includes a major rewrite of the NFSv4 state code, which has
always depended on a single mutex. As an example, open creates are no
longer serialized, fixing a performance regression on NFSv3->NFSv4
upgrades. Thanks to Jeff, Trond, and Benny, and to Christoph for
review.
Also some RDMA fixes from Chuck Lever and Steve Wise, and
miscellaneous fixes from Kinglong Mee and others"
* 'for-3.17' of git://linux-nfs.org/~bfields/linux: (167 commits)
svcrdma: remove rdma_create_qp() failure recovery logic
nfsd: add some comments to the nfsd4 object definitions
nfsd: remove the client_mutex and the nfs4_lock/unlock_state wrappers
nfsd: remove nfs4_lock_state: nfs4_state_shutdown_net
nfsd: remove nfs4_lock_state: nfs4_laundromat
nfsd: Remove nfs4_lock_state(): reclaim_complete()
nfsd: Remove nfs4_lock_state(): setclientid, setclientid_confirm, renew
nfsd: Remove nfs4_lock_state(): exchange_id, create/destroy_session()
nfsd: Remove nfs4_lock_state(): nfsd4_open and nfsd4_open_confirm
nfsd: Remove nfs4_lock_state(): nfsd4_delegreturn()
nfsd: Remove nfs4_lock_state(): nfsd4_open_downgrade + nfsd4_close
nfsd: Remove nfs4_lock_state(): nfsd4_lock/locku/lockt()
nfsd: Remove nfs4_lock_state(): nfsd4_release_lockowner
nfsd: Remove nfs4_lock_state(): nfsd4_test_stateid/nfsd4_free_stateid
nfsd: Remove nfs4_lock_state(): nfs4_preprocess_stateid_op()
nfsd: remove old fault injection infrastructure
nfsd: add more granular locking to *_delegations fault injectors
nfsd: add more granular locking to forget_openowners fault injector
nfsd: add more granular locking to forget_locks fault injector
nfsd: add a list_head arg to nfsd_foreach_client_lock
...
In svc_rdma_accept(), if rdma_create_qp() fails, there is useless
logic to try and call rdma_create_qp() again with reduced sge depths.
The assumption, I guess, was that perhaps the initial sge depths
chosen were too big. However they initial depths are selected based
on the rdma device attribute max_sge returned from ib_query_device().
If rdma_create_qp() fails, it would not be because the max_send_sge and
max_recv_sge values passed in exceed the device's max. So just remove
this code.
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
current_cred() can only be changed by 'current', and
cred->group_info is never changed. If a new group_info is
needed, a new 'cred' is created.
Consequently it is always safe to access
current_cred()->group_info
without taking any further references.
So drop the refcounting and the incorrect rcu_dereference().
Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
The new flag RPCAUTH_LOOKUP_RCU to credential lookup avoids locking,
does not take a reference on the returned credential, and returns
-ECHILD if a simple lookup was not possible.
The returned value can only be used within an rcu_read_lock protected
region.
The main user of this is the new rpc_lookup_cred_nonblock() which
returns a pointer to the current credential which is only rcu-safe (no
ref-count held), and might return -ECHILD if allocation was required.
Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Fix the endianness handling in gss_wrap_kerberos_v1 and drop the memset
call there in favor of setting the filler bytes directly.
In gss_wrap_kerberos_v2, get rid of the "ec" variable which is always
zero, and drop the endianness conversion of 0. Sparse handles 0 as a
special case, so it's not necessary.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Use u16 pointer in setup_token and setup_token_v2. None of the fields
are actually handled as __be16, so this simplifies the code a bit. Also
get rid of some unneeded pointer increments.
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
The handling of the gc_ctx pointer only seems to be partially RCU-safe.
The assignment and freeing are done using RCU, but many places in the
code seem to dereference that pointer without proper RCU safeguards.
Fix them to use rcu_dereference and to rcu_read_lock/unlock, and to
properly handle the case where the pointer is NULL.
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Paul McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: Jeff Layton <jlayton@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
* 'nfs-rdma' of git://git.linux-nfs.org/projects/anna/nfs-rdma: (916 commits)
xprtrdma: Handle additional connection events
xprtrdma: Remove RPCRDMA_PERSISTENT_REGISTRATION macro
xprtrdma: Make rpcrdma_ep_disconnect() return void
xprtrdma: Schedule reply tasklet once per upcall
xprtrdma: Allocate each struct rpcrdma_mw separately
xprtrdma: Rename frmr_wr
xprtrdma: Disable completions for LOCAL_INV Work Requests
xprtrdma: Disable completions for FAST_REG_MR Work Requests
xprtrdma: Don't post a LOCAL_INV in rpcrdma_register_frmr_external()
xprtrdma: Reset FRMRs after a flushed LOCAL_INV Work Request
xprtrdma: Reset FRMRs when FAST_REG_MR is flushed by a disconnect
xprtrdma: Properly handle exhaustion of the rb_mws list
xprtrdma: Chain together all MWs in same buffer pool
xprtrdma: Back off rkey when FAST_REG_MR fails
xprtrdma: Unclutter struct rpcrdma_mr_seg
xprtrdma: Don't invalidate FRMRs if registration fails
xprtrdma: On disconnect, don't ignore pending CQEs
xprtrdma: Update rkeys after transport reconnect
xprtrdma: Limit data payload size for ALLPHYSICAL
xprtrdma: Protect ia->ri_id when unmapping/invalidating MRs
...
In some cases where the credentials are not often reused, we may want
to limit their total number just in order to make the negative lookups
in the hash table more manageable.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Commit 38ca83a5 added RDMA_CM_EVENT_TIMEWAIT_EXIT. But that status
is relevant only for consumers that re-use their QPs on new
connections. xprtrdma creates a fresh QP on reconnection, so that
event should be explicitly ignored.
Squelch the alarming "unexpected CM event" message.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Clean up.
RPCRDMA_PERSISTENT_REGISTRATION was a compile-time switch between
RPCRDMA_REGISTER mode and RPCRDMA_ALLPHYSICAL mode. Since
RPCRDMA_REGISTER has been removed, there's no need for the extra
conditional compilation.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Clean up: The return code is used only for dprintk's that are
already redundant.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Minor optimization: grab rpcrdma_tk_lock_g and disable hard IRQs
just once after clearing the receive completion queue.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Currently rpcrdma_buffer_create() allocates struct rpcrdma_mw's as
a single contiguous area of memory. It amounts to quite a bit of
memory, and there's no requirement for these to be carved from a
single piece of contiguous memory.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Clean up: Name frmr_wr after the opcode of the Work Request,
consistent with the send and local invalidation paths.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Instead of relying on a completion to change the state of an FRMR
to FRMR_IS_INVALID, set it in advance. If an error occurs, a completion
will fire anyway and mark the FRMR FRMR_IS_STALE.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Instead of relying on a completion to change the state of an FRMR
to FRMR_IS_VALID, set it in advance. If an error occurs, a completion
will fire anyway and mark the FRMR FRMR_IS_STALE.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Any FRMR arriving in rpcrdma_register_frmr_external() is now
guaranteed to be either invalid, or to be targeted by a queued
LOCAL_INV that will invalidate it before the adapter processes
the FAST_REG_MR being built here.
The problem with current arrangement of chaining a LOCAL_INV to the
FAST_REG_MR is that if the transport is not connected, the LOCAL_INV
is flushed and the FAST_REG_MR is flushed. This leaves the FRMR
valid with the old rkey. But rpcrdma_register_frmr_external() has
already bumped the in-memory rkey.
Next time through rpcrdma_register_frmr_external(), a LOCAL_INV and
FAST_REG_MR is attempted again because the FRMR is still valid. But
the rkey no longer matches the hardware's rkey, and a memory
management operation error occurs.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
When a LOCAL_INV Work Request is flushed, it leaves an FRMR in the
VALID state. This FRMR can be returned by rpcrdma_buffer_get(), and
must be knocked down in rpcrdma_register_frmr_external() before it
can be re-used.
Instead, capture these in rpcrdma_buffer_get(), and reset them.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
FAST_REG_MR Work Requests update a Memory Region's rkey. Rkey's are
used to block unwanted access to the memory controlled by an MR. The
rkey is passed to the receiver (the NFS server, in our case), and is
also used by xprtrdma to invalidate the MR when the RPC is complete.
When a FAST_REG_MR Work Request is flushed after a transport
disconnect, xprtrdma cannot tell whether the WR actually hit the
adapter or not. So it is indeterminant at that point whether the
existing rkey is still valid.
After the transport connection is re-established, the next
FAST_REG_MR or LOCAL_INV Work Request against that MR can sometimes
fail because the rkey value does not match what xprtrdma expects.
The only reliable way to recover in this case is to deregister and
register the MR before it is used again. These operations can be
done only in a process context, so handle it in the transport
connect worker.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
If the rb_mws list is exhausted, clean up and return NULL so that
call_allocate() will delay and try again.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
During connection loss recovery, need to visit every MW in a
buffer pool. Any MW that is in use by an RPC will not be on the
rb_mws list.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
If posting a FAST_REG_MR Work Reqeust fails, revert the rkey update
to avoid subsequent IB_WC_MW_BIND_ERR completions.
Suggested-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Clean ups:
- make it obvious that the rl_mw field is a pointer -- allocated
separately, not as part of struct rpcrdma_mr_seg
- promote "struct {} frmr;" to a named type
- promote the state enum to a named type
- name the MW state field the same way other fields in
rpcrdma_mw are named
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
If FRMR registration fails, it's likely to transition the QP to the
error state. Or, registration may have failed because the QP is
_already_ in ERROR.
Thus calling rpcrdma_deregister_external() in
rpcrdma_create_chunks() is useless in FRMR mode: the LOCAL_INVs just
get flushed.
It is safe to leave existing registrations: when FRMR registration
is tried again, rpcrdma_register_frmr_external() checks if each FRMR
is already/still VALID, and knocks it down first if it is.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
xprtrdma is currently throwing away queued completions during
a reconnect. RPC replies posted just before connection loss, or
successful completions that change the state of an FRMR, can be
missed.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Various reports of:
rpcrdma_qp_async_error_upcall: QP error 3 on device mlx4_0
ep ffff8800bfd3e848
Ensure that rkeys in already-marshalled RPC/RDMA headers are
refreshed after the QP has been replaced by a reconnect.
BugLink: https://bugzilla.linux-nfs.org/show_bug.cgi?id=249
Suggested-by: Selvin Xavier <Selvin.Xavier@Emulex.Com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
When the client uses physical memory registration, each page in the
payload gets its own array entry in the RPC/RDMA header's chunk list.
Therefore, don't advertise a maximum payload size that would require
more array entries than can fit in the RPC buffer where RPC/RDMA
headers are built.
BugLink: https://bugzilla.linux-nfs.org/show_bug.cgi?id=248
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Ensure ia->ri_id remains valid while invoking dma_unmap_page() or
posting LOCAL_INV during a transport reconnect. Otherwise,
ia->ri_id->device or ia->ri_id->qp is NULL, which triggers a panic.
BugLink: https://bugzilla.linux-nfs.org/show_bug.cgi?id=259
Fixes: ec62f40 'xprtrdma: Ensure ia->ri_id->qp is not NULL when reconnecting'
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
seg1->mr_nsegs is not yet initialized when it is used to unmap
segments during an error exit. Use the same unmapping logic for
all error exits.
"if (frmr_wr.wr.fast_reg.length < len) {" used to be a BUG_ON check.
The broken code will never be executed under normal operation.
Fixes: c977dea (xprtrdma: Remove BUG_ON() call sites)
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Tested-by: Shirley Ma <shirley.ma@oracle.com>
Tested-by: Devesh Sharma <devesh.sharma@emulex.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
If requests are queued in the socket inbuffer waiting for an
svc_tcp_has_wspace() requirement to be satisfied, then we do not want
to clear the SOCK_NOSPACE flag until we've satisfied that requirement.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>
Ensure that all calls to svc_xprt_enqueue() except svc_xprt_received()
check the value of XPT_BUSY, before attempting to grab spinlocks etc.
This is to avoid situations such as the following "perf" trace,
which shows heavy contention on the pool spinlock:
54.15% nfsd [kernel.kallsyms] [k] _raw_spin_lock_bh
|
--- _raw_spin_lock_bh
|
|--71.43%-- svc_xprt_enqueue
| |
| |--50.31%-- svc_reserve
| |
| |--31.35%-- svc_xprt_received
| |
| |--18.34%-- svc_tcp_data_ready
...
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Signed-off-by: J. Bruce Fields <bfields@redhat.com>