| 539431a437d2e5d6d94016184dfc0aab263c01e1 |
|
29-Jul-2014 |
Chuck Lever <chuck.lever@oracle.com> |
xprtrdma: Don't invalidate FRMRs if registration fails
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>
|
| 6ab59945f292a5c6cbc4a6c2011f1a732a116af2 |
|
29-Jul-2014 |
Chuck Lever <chuck.lever@oracle.com> |
xprtrdma: Update rkeys after transport reconnect
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refreshed after the QP has been replaced by a reconnect.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>
|
| c93c62231cf55df4a26bd08937efeea97e6fc5e8 |
|
28-May-2014 |
Chuck Lever <chuck.lever@oracle.com> |
xprtrdma: Disconnect on registration failure
If rpcrdma_register_external() fails during request marshaling, the
current RPC request is killed. Instead, this RPC should be retried
after reconnecting the transport instance.
The most likely reason for registration failure with FRMR is a
failed post_send, which would be due to a remote transport
disconnect or memory exhaustion. These issues can be recovered
by a retry.
Problems encountered in the marshaling logic itself will not be
corrected by trying again, so these should still kill a request.
Now that we've added a clean exit for marshaling errors, take the
opportunity to defang some BUG_ON's.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
|
| e7ce710a8802351bd4118c5d6136c1d850f67cf9 |
|
28-May-2014 |
Chuck Lever <chuck.lever@oracle.com> |
xprtrdma: Avoid deadlock when credit window is reset
Update the cwnd while processing the server's reply. Otherwise the
next task on the xprt_sending queue is still subject to the old
credit window. Currently, no task is awoken if the old congestion
window is still exceeded, even if the new window is larger, and a
deadlock results.
This is an issue during a transport reconnect. Servers don't
normally shrink the credit window, but the client does reset it to
1 when reconnecting so the server can safely grow it again.
As a minor optimization, remove the hack of grabbing the initial
cwnd size (which happens to be RPC_CWNDSCALE) and using that value
as the congestion scaling factor. The scaling value is invariant,
and we are better off without the multiplication operation.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
|
| 18906972aa1103c07869c9b43860a52e0e27e8e5 |
|
28-May-2014 |
Chuck Lever <chuck.lever@oracle.com> |
xprtrdma: Reset connection timeout after successful reconnect
If the new connection is able to make forward progress, reset the
re-establish timeout. Otherwise it keeps growing even if disconnect
events are rare.
The same behavior as TCP is adopted: reconnect immediately if the
transport instance has been able to make some forward progress.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
|
| 196c69989d84ab902bbe545f7bd8ce78ee74dac4 |
|
28-May-2014 |
Shirley Ma <shirley.ma@oracle.com> |
xprtrdma: Allocate missing pagelist
GETACL relies on transport layer to alloc memory for reply buffer.
However xprtrdma assumes that the reply buffer (pagelist) has been
pre-allocated in upper layer. This problem was reported by IOL OFA lab
test on PPC.
Signed-off-by: Shirley Ma <shirley.ma@oracle.com>
Reviewed-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Edward Mossman <emossman@iol.unh.edu>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
|
| 13c9ff8f673862b69e795ea99a237b461c557eb3 |
|
28-May-2014 |
Chuck Lever <chuck.lever@oracle.com> |
xprtrdma: Simplify rpcrdma_deregister_external() synopsis
Clean up: All remaining callers of rpcrdma_deregister_external()
pass NULL as the last argument, so remove that argument.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
|
| 0ac531c1832318efa3dc3d723e356a7e09330e80 |
|
28-May-2014 |
Chuck Lever <chuck.lever@oracle.com> |
xprtrdma: Remove REGISTER memory registration mode
All kernel RDMA providers except amso1100 support either MTHCAFMR
or FRMR, both of which are faster than REGISTER. amso1100 can
continue to use ALLPHYSICAL.
The only other ULP consumer in the kernel that uses the reg_phys_mr
verb is Lustre.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
|
| b45ccfd25d506e83d9ecf93d0ac7edf031d35d2f |
|
28-May-2014 |
Chuck Lever <chuck.lever@oracle.com> |
xprtrdma: Remove MEMWINDOWS registration modes
The MEMWINDOWS and MEMWINDOWS_ASYNC memory registration modes were
intended as stop-gap modes before the introduction of FRMR. They
are now considered obsolete.
MEMWINDOWS_ASYNC is also considered unsafe because it can leave
client memory registered and exposed for an indeterminant time after
each I/O.
At this point, the MEMWINDOWS modes add needless complexity, so
remove them.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
|
| 03ff8821eb5ed168792667cfc3ddff903e97af99 |
|
28-May-2014 |
Chuck Lever <chuck.lever@oracle.com> |
xprtrdma: Remove BOUNCEBUFFERS memory registration mode
Clean up: This memory registration mode is slow and was never
meant for use in production environments. Remove it to reduce
implementation complexity.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
|
| 254f91e2fa1f4cc18fd2eb9d5481888ffe126d5b |
|
28-May-2014 |
Chuck Lever <chuck.lever@oracle.com> |
xprtrdma: RPC/RDMA must invoke xprt_wake_pending_tasks() in process context
An IB provider can invoke rpcrdma_conn_func() in an IRQ context,
thus rpcrdma_conn_func() cannot be allowed to directly invoke
generic RPC functions like xprt_wake_pending_tasks().
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Tested-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
|
| 0fc6c4e7bb287148eb5e949efd89327929d4841d |
|
28-May-2014 |
Steve Wise <swise@opengridcomputing.com> |
xprtrdma: mind the device's max fast register page list depth
Some rdma devices don't support a fast register page list depth of
at least RPCRDMA_MAX_DATA_SEGS. So xprtrdma needs to chunk its fast
register regions according to the minimum of the device max supported
depth or RPCRDMA_MAX_DATA_SEGS.
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Reviewed-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
|
| 2b7bbc963da8d076f263574af4138b5df2e1581f |
|
12-Mar-2014 |
Chuck Lever <chuck.lever@oracle.com> |
SUNRPC: Fix large reads on NFS/RDMA
After commit a11a2bf4, "SUNRPC: Optimise away unnecessary data moves
in xdr_align_pages", Thu Aug 2 13:21:43 2012, READs larger than a
few hundred bytes via NFS/RDMA no longer work. This commit exposed
a long-standing bug in rpcrdma_inline_fixup().
I reproduce this with an rsize=4096 mount using the cthon04 basic
tests. Test 5 fails with an EIO error.
For my reproducer, kernel log shows:
NFS: server cheating in read reply: count 4096 > recvd 0
rpcrdma_inline_fixup() is zeroing the xdr_stream::page_len field,
and xdr_align_pages() is now returning that value to the READ XDR
decoder function.
That field is set up by xdr_inline_pages() by the READ XDR encoder
function. As far as I can tell, it is supposed to be left alone
after that, as it describes the dimensions of the reply xdr_stream,
not the contents of that stream.
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=68391
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
|
| a4f0835c604f80f945ab3e72ffd00547145c4b2b |
|
08-Jan-2013 |
Trond Myklebust <Trond.Myklebust@netapp.com> |
SUNRPC: Eliminate task->tk_xprt accesses that bypass rcu_dereference()
tk_xprt is just a shortcut for tk_client->cl_xprt, however cl_xprt is
defined as an __rcu variable. Replace dereferences of tk_xprt with
non-rcu dereferences where it is safe to do so.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
|
| 4a6862b3649d705bf41a36e3c7943d0322a9ee27 |
|
20-Feb-2012 |
Tom Tucker <tom@ogc.us> |
xprtrdma: The transport should not bug-check when a dup reply is received
The client side RDMA transport will bug check if it receives a duplicate
reply, instead we should simply drop the duplicate reply.
Signed-off-by: Tom Tucker <tom@ogc.us>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
|
| b85417860172ff693dc115d7999805fc240cec1c |
|
25-Nov-2011 |
Cong Wang <amwang@redhat.com> |
sunrpc: remove the second argument of k[un]map_atomic()
Signed-off-by: Cong Wang <amwang@redhat.com>
|
| bd7ea31b9e8a342be76e0fe8d638343886c2d8c5 |
|
09-Feb-2011 |
Tom Tucker <tom@ogc.us> |
RPCRDMA: Fix to XDR page base interpretation in marshalling logic.
The RPCRDMA marshalling logic assumed that xdr->page_base was an
offset into the first page of xdr->page_list. It is in fact an
offset into the xdr->page_list itself, that is, it selects the
first page in the page_list and the offset into that page.
The symptom depended in part on the rpc_memreg_strategy, if it was
FRMR, or some other one-shot mapping mode, the connection would get
torn down on a base and bounds error. When the badly marshalled RPC
was retransmitted it would reconnect, get the error, and tear down the
connection again in a loop forever. This resulted in a hung-mount. For
the other modes, it would result in silent data corruption. This bug is
most easily reproduced by writing more data than the filesystem
has space for.
This fix corrects the page_base assumption and otherwise simplifies
the iov mapping logic.
Signed-off-by: Tom Tucker <tom@ogc.us>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
|
| 15cdc644b268a9a9ce73ce0b153129222c254b7b |
|
11-Aug-2010 |
Tom Tucker <tom@ogc.us> |
rpcrdma: Fix SQ size calculation when memreg is FRMR
This patch updates the computation to include the worst case situation
where three FRMR are required to map a single RPC REQ.
Signed-off-by: Tom Tucker <tom@ogc.us>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
|
| b38ab40ad58c1fc43ea590d6342f6a6763ac8fb6 |
|
11-Mar-2009 |
Tom Talpey <tmtalpey@gmail.com> |
XPRTRDMA: correct an rpc/rdma inline send marshaling error
Certain client rpc's which contain both lengthy page-contained
metadata and a non-empty xdr_tail buffer require careful handling
to avoid overlapped memory copying. Rearranging of existing rpcrdma
marshaling code avoids it; this fixes an NFSv4 symlink creation error
detected with connectathon basic/test8 to multiple servers.
Signed-off-by: Tom Talpey <tmtalpey@gmail.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
|
| 5f37d561e0f0cd98017c389cbc22080290f11c3c |
|
09-Oct-2008 |
Tom Talpey <talpey@netapp.com> |
RPC/RDMA: reformat a debug printk to keep lines together.
The send marshaling code split a particular dprintk across two
lines, which makes it hard to extract from logfiles.
Signed-off-by: Tom Talpey <talpey@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
|
| 926449ba66ce2a45c619bbe755b00d6bdbf0d83e |
|
09-Oct-2008 |
Tom Talpey <talpey@netapp.com> |
RPC/RDMA: return a consistent error, when connect fails.
The xprt_connect call path does not expect such errors as ECONNREFUSED
to be returned from failed transport connection attempts, otherwise it
translates them to EIO and signals fatal errors. For example, mount.nfs
prints simply "internal error". Translate all such errors to ENOTCONN
from RPC/RDMA to match sockets behavior.
Signed-off-by: Tom Talpey <talpey@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
|
| 9191ca3b381b15b9a88785a8ae2fa4db8e553b0c |
|
09-Oct-2008 |
Tom Talpey <talpey@netapp.com> |
RPC/RDMA: adhere to protocol for unpadded client trailing write chunks.
The RPC/RDMA protocol allows clients and servers to avoid RDMA
operations for data which is purely the result of XDR padding.
On the client, automatically insert the necessary padding for
such server replies, and optionally don't marshal such chunks.
Signed-off-by: Tom Talpey <talpey@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
|
| 575448bd36208f99fe0dd554a43518d798966740 |
|
09-Oct-2008 |
Tom Talpey <talpey@netapp.com> |
RPC/RDMA: suppress retransmit on RPC/RDMA clients.
An RPC/RDMA client cannot retransmit on an unbroken connection,
doing so violates its flow control with the server.
Signed-off-by: Tom Talpey <talpey@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
|
| 6067804047b64dde89f4f133fc7eba48ee44107d |
|
21-Sep-2008 |
Arnaldo Carvalho de Melo <acme@redhat.com> |
net: Use hton[sl]() instead of __constant_hton[sl]() where applicable
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| d4b37ff73540ab90bee57b882a10b21e2f97939f |
|
26-Oct-2007 |
Chuck Lever <chuck.lever@oracle.com> |
SUNRPC: Fix an unnecessary implicit type cast in rpcrdma_count_chunks()
Nit: rl_nchunks is an unsigned integer, so pass it into
rpcrdma_count_chunks() via an unsigned integer argument. This eliminates
a harmless mixed sign comparison in rpcrdma_count_chunks()
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Cc: Thomas Talpey <Thomas.Talpey@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
|
| 2a428b2b8fe2c270a5889086ebe3ab914e3ea7d8 |
|
26-Oct-2007 |
Chuck Lever <chuck.lever@oracle.com> |
SUNRPC: Prevent mixed sign comparisons in rpcrdma_convert_iovs()
Keep the type of the buffer position the same during iovec conversion to
reduce the likelihood of unexpected results from comparisons and length
computations.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Cc: Thomas Talpey <Thomas.Talpey@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
|
| 8d614434ab77b440b69e66a9bd44e46e7194c34a |
|
11-Dec-2007 |
YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org> |
[SUNRPC]: Use htonl() where appropriate.
Signed-off-by: YOSHIFUJI Hideaki <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| 50e1092b3a119bb4660bb6bd2e1749dc2d8ac62e |
|
10-Dec-2007 |
James Lentini <jlentini@netapp.com> |
SUNRPC xprtrdma: fix XDR tail buf marshalling for all ops
rpcrdma_convert_iovs is passed an xdr_buf representing either an RPC
request or an RPC reply. In the case of a request, several
calculations and tests involving pos are unnecessary. In the case of a
reply, several calculations and tests involving pos are incorrect (the
code tests pos against the reply xdr buf's len field, which is always
0 at the time rpcrdma_convert_iovs is executed). This change removes
the incorrect/unnecessary calculations and tests involving pos.
This fixes an observed problem when reading certain file sizes over
NFS/RDMA.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Signed-off-by: Tom Talpey <talpey@netapp.com>
Signed-off-by: James Lentini <jlentini@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
|
| e08a132b0ef3cf89dfbf1dea2c6248ea624bdcd7 |
|
30-Oct-2007 |
Stephen Rothwell <sfr@canb.auug.org.au> |
[SUNRPC] rpc_rdma: we need to cast u64 to unsigned long long for printing
as some architectures have unsigned long for u64.
net/sunrpc/xprtrdma/rpc_rdma.c: In function 'rpcrdma_create_chunks':
net/sunrpc/xprtrdma/rpc_rdma.c:222: warning: format '%llx' expects type 'long long unsigned int', but argument 4 has type 'u64'
net/sunrpc/xprtrdma/rpc_rdma.c:234: warning: format '%llx' expects type 'long long unsigned int', but argument 5 has type 'u64'
net/sunrpc/xprtrdma/rpc_rdma.c: In function 'rpcrdma_count_chunks':
net/sunrpc/xprtrdma/rpc_rdma.c:577: warning: format '%llx' expects type 'long long unsigned int', but argument 4 has type 'u64
Noticed on PowerPC pseries_defconfig build.
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
| 2d8a972661832719931b0dd5b80e97215cb93d94 |
|
29-Oct-2007 |
Al Viro <viro@ftp.linux.org.uk> |
SUNRPC endianness annotations
rpcrdma stuff lacks endianness annotations for on-the-wire data.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
| e96018280cb36210f4c69663561825114a57e7e1 |
|
10-Sep-2007 |
\"Talpey, Thomas\ <Thomas.Talpey@netapp.com> |
RPCRDMA: rpc rdma protocol implementation
This implements the marshaling and unmarshaling of the rpcrdma transport
headers. Connection management is also addressed.
Signed-off-by: Tom Talpey <talpey@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
|
| f58851e6b0f148fb4b2a1c6f70beb2f125863c0f |
|
10-Sep-2007 |
\"Talpey, Thomas\ <Thomas.Talpey@netapp.com> |
RPCRDMA: rpc rdma transport switch
This implements the configuration and building of the core transport
switch implementation of the rpcrdma transport. Stubs are provided for
the rpcrdma protocol handling, and the infiniband/iwarp verbs interface.
These are provided in following patches.
Signed-off-by: Tom Talpey <talpey@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
|