* 'for-2.6.28' of git://linux-nfs.org/~bfields/linux: (59 commits)
svcrdma: Fix IRD/ORD polarity
svcrdma: Update svc_rdma_send_error to use DMA LKEY
svcrdma: Modify the RPC reply path to use FRMR when available
svcrdma: Modify the RPC recv path to use FRMR when available
svcrdma: Add support to svc_rdma_send to handle chained WR
svcrdma: Modify post recv path to use local dma key
svcrdma: Add a service to register a Fast Reg MR with the device
svcrdma: Query device for Fast Reg support during connection setup
svcrdma: Add FRMR get/put services
NLM: Remove unused argument from svc_addsock() function
NLM: Remove "proto" argument from lockd_up()
NLM: Always start both UDP and TCP listeners
lockd: Remove unused fields in the nlm_reboot structure
lockd: Add helper to sanity check incoming NOTIFY requests
lockd: change nlmclnt_grant() to take a "struct sockaddr *"
lockd: Adjust nlmsvc_lookup_host() to accomodate AF_INET6 addresses
lockd: Adjust nlmclnt_lookup_host() signature to accomodate non-AF_INET
lockd: Support non-AF_INET addresses in nlm_lookup_host()
NLM: Convert nlm_lookup_host() to use a single argument
svcrdma: Add Fast Reg MR Data Types
...
The RPC/RDMA connection logic could return early from reconnection
attempts, leading to additional spurious retries.
Signed-off-by: Tom Talpey <talpey@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
The RPC/RDMA code had a constant 5-second reconnect backoff, and
always performed it, even when re-establishing a connection to a
server after the RPC layer closed it due to being idle. Make it
an geometric backoff (up to 30 seconds), and don't delay idle
reconnect.
Signed-off-by: Tom Talpey <talpey@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
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>
Add defensive timeouts to wait_for_completion() calls in RDMA
address resolution, and make them interruptible. Fix the timeout
units to milliseconds (formerly jiffies) and move to private header.
Signed-off-by: Tom Talpey <talpey@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
The RPC/RDMA code can leak RDMA connection manager endpoints in
certain error cases on connect. Don't signal unwanted events,
and be certain to destroy any allocated qp.
Signed-off-by: Tom Talpey <talpey@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
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>
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>
RDMA disconnects yield an upcall from the RDMA connection manager,
which can race with rpc transport close, e.g. on ^C of a mount.
Ensure any rdma cm_id and qp are fully destroyed before continuing.
Signed-off-by: Tom Talpey <talpey@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
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>
This logic sets the connection parameter that configures the local device
and informs the remote peer how many concurrent incoming RDMA_READ
requests are supported. The original logic didn't really do what was
intended for two reasons:
- The max number supported by the device is typically smaller than
any one factor in the calculation used, and
- The field in the connection parameter structure where the value is
stored is a u8 and always overflows for the default settings.
So what really happens is the value requested for responder resources
is the left over 8 bits from the "desired value". If the desired value
happened to be a multiple of 256, the result was zero and it wouldn't
connect at all.
Given the above and the fact that max_requests is almost always larger
than the max responder resources supported by the adapter, this patch
simplifies this logic and simply requests the max supported by the device,
subject to a reasonable limit.
This bug was found by Jim Schutt at Sandia.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Acked-by: Tom Talpey <talpey@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Configure, detect and use "fastreg" support from IB/iWARP verbs
layer to perform RPC/RDMA memory registration.
Make FRMR the default memreg mode (will fall back if not supported
by the selected RDMA adapter).
This allows full and optimal operation over the cxgb3 adapter, and others.
Signed-off-by: Tom Talpey <talpey@netapp.com>
Acked-by: Tom Tucker <tom@opengridcomputing.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
At transport creation, check for, and use, any local dma lkey.
Then, check that the selected memory registration mode is in fact
supported by the RDMA adapter selected for the mount. Fall back
to best alternative if not.
Signed-off-by: Tom Talpey <talpey@netapp.com>
Acked-by: Tom Tucker <tom@opengridcomputing.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Refactor the memory registration and deregistration routines.
This saves stack space, makes the code more readable and prepares
to add the new FRMR registration methods.
Signed-off-by: Tom Talpey <talpey@netapp.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
The inititator/responder resources in the event have been swapped. They
no represent what the local peer would set their values to in order to
match the peer. Note that iWARP does not exchange these on the wire and
the provider is simply putting in the local device max.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Update the svc_rdma_send_error code to use the DMA LKEY which is valid
regardless of the memory registration strategy in use.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Use FRMR to map local RPC reply data. This allows RDMA_WRITE to send reply
data using a single WR. The FRMR is invalidated by linking the LOCAL_INV WR
to the RDMA_SEND message used to complete the reply.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
RPCRDMA requests that specify a read-list are fetched with RDMA_READ. Using
an FRMR to map the data sink improves NFSRDMA security on transports that
place the RDMA_READ data sink LKEY on the wire because the valid lifetime
of the MR is only the duration of the RDMA_READ. The LKEY is invalidated
when the last RDMA_READ WR completes.
Mapping the data sink also allows for very large amounts to data to be
fetched with a single WR, so if the client is also using FRMR, the entire
RPC read-list can be fetched with a single WR.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
WR can be submitted as linked lists of WR. Update the svc_rdma_send
routine to handle WR chains. This will be used to submit a WR that
uses an FRMR with another WR that invalidates the FRMR.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Update the svc_rdma_post_recv routine to use the adapter's global LKEY
instead of sc_phys_mr which is only valid when using a DMA MR.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Fast Reg MR introduces a new WR type. Add a service to register the
region with the adapter and update the completion handling to support
completions with a NULL WR context.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Query the device capabilities in the svc_rdma_accept function to determine
what advanced memory management capabilities are supported by the device.
Based on the query, select the most secure model available given the
requirements of the transport and capabilities of the adapter.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Add services for the allocating, freeing, and unmapping Fast Reg MR. These
services will be used by the transport connection setup, send and receive
routines.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
RDMA_READ completions are kept on a separate queue from the general
I/O request queue. Since a separate lock is used to protect the RDMA_READ
completion queue, a race exists between the dto_tasklet and the
svc_rdma_recvfrom thread where the dto_tasklet sets the XPT_DATA
bit and adds I/O to the read-completion queue. Concurrently, the
recvfrom thread checks the generic queue, finds it empty and resets
the XPT_DATA bit. A subsequent svc_xprt_enqueue will fail to enqueue
the transport for I/O and cause the transport to "stall".
The fix is to protect both lists with the same lock and set the XPT_DATA
bit with this lock held.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Add per-device dma_mapping_ops support for CONFIG_X86_64 as POWER
architecture does:
This enables us to cleanly fix the Calgary IOMMU issue that some devices
are not behind the IOMMU (http://lkml.org/lkml/2008/5/8/423).
I think that per-device dma_mapping_ops support would be also helpful for
KVM people to support PCI passthrough but Andi thinks that this makes it
difficult to support the PCI passthrough (see the above thread). So I
CC'ed this to KVM camp. Comments are appreciated.
A pointer to dma_mapping_ops to struct dev_archdata is added. If the
pointer is non NULL, DMA operations in asm/dma-mapping.h use it. If it's
NULL, the system-wide dma_ops pointer is used as before.
If it's useful for KVM people, I plan to implement a mechanism to register
a hook called when a new pci (or dma capable) device is created (it works
with hot plugging). It enables IOMMUs to set up an appropriate
dma_mapping_ops per device.
The major obstacle is that dma_mapping_error doesn't take a pointer to the
device unlike other DMA operations. So x86 can't have dma_mapping_ops per
device. Note all the POWER IOMMUs use the same dma_mapping_error function
so this is not a problem for POWER but x86 IOMMUs use different
dma_mapping_error functions.
The first patch adds the device argument to dma_mapping_error. The patch
is trivial but large since it touches lots of drivers and dma-mapping.h in
all the architecture.
This patch:
dma_mapping_error() doesn't take a pointer to the device unlike other DMA
operations. So we can't have dma_mapping_ops per device.
Note that POWER already has dma_mapping_ops per device but all the POWER
IOMMUs use the same dma_mapping_error function. x86 IOMMUs use device
argument.
[akpm@linux-foundation.org: fix sge]
[akpm@linux-foundation.org: fix svc_rdma]
[akpm@linux-foundation.org: build fix]
[akpm@linux-foundation.org: fix bnx2x]
[akpm@linux-foundation.org: fix s2io]
[akpm@linux-foundation.org: fix pasemi_mac]
[akpm@linux-foundation.org: fix sdhci]
[akpm@linux-foundation.org: build fix]
[akpm@linux-foundation.org: fix sparc]
[akpm@linux-foundation.org: fix ibmvscsi]
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Muli Ben-Yehuda <muli@il.ibm.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Avi Kivity <avi@qumranet.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Change the WR context pool to be shared across mount points. This
reduces the RDMA transport memory footprint significantly since
idle mounts don't consume WR context memory.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Create a kmem cache to hold WR contexts. Next we will convert
the WR context get and put services to use this kmem cache.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
When adapters have differing IRD limits, the RDMA transport will fail to
connect properly. The RDMA transport should use the client's advertised
inbound read limit when computing its outbound read limit. For iWARP
transports, there is currently no standard for exchanging IRD/ORD
during connection establishment so the 'responder_resources' field in the
connect event is the local device's limit. The RDMA transport can be
configured to use a smaller ORD by writing the desired number to the
/proc/sys/sunrpc/svc_rdma/max_outbound_read_requests file.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
At the time __svc_rdma_free is called, we are guaranteed that all references
to this transport are gone. There is, therefore, no need to protect the
resource lists with a spin lock.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Add a dma map count in order to verify that all DMA mapping resources
have been freed when the transport is closed.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Separate DMA unmap from context destruction and perform DMA unmapping
in the SQ/RQ CQ reap functions. This is necessary to support software
based RDMA implementations that actually copy the data in their
ib_dma_unmap callback functions and architectures that don't have
cache coherent I/O busses.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Modify the RDMA_READ processing to use the reply and chunk list mapping data
types. Also add a special purpose 'hdr_count' field in in the context to hold
the header page count instead of overloading the SGE length field and
corrupting the DMA map length.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Use the new svc_rdma_req_map data type for mapping the client side memory
to the server side memory. Move the DMA mapping to the context pointed to
by each WR individually so that it is unmapped after the WR completes.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Create a new data structure to hold the remote client address space
to local server address space mapping.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
A RDMA read-list cannot contain more elements than RPCSVC_MAXPAGES or
it will overflow the DTO context. Verify this when processing the
protocol header.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
The svc_rdma_send_error function is called when an RPCRDMA protocol
error is detected. This function attempts to post an error reply message.
Since an error posting to a transport in error is ignored, change
the return type to void.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
This race was found by inspection. Messages can be received from the peer
immediately following the rdma_accept call, however, the CQ have not yet
been armed and the transport address has not yet been set.
Set the transport address in the connect request handler and arm the CQ
prior to calling rdma_accept.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
The rdma_read_complete function needs to copy the rqstp transport address
from the transport. Failure to do so can result in using the wrong
authentication method for the RPC or bug checking if the rqstp address
is not valid.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Use the ib_verbs version of the dma_unmap service in the
svc_rdma_put_context function. This should support providers
using software rdma.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
When the transport is closing, the DTO tasklet may queue data
that never gets processed. Clean up resources associated with
this I/O.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Move the destruction of the QP and CM_ID to the free path so that the
QP cleanup code doesn't race with the dto_tasklet handling flushed WR.
The QP reference is not needed because we now have a reference for
every WR.
Also add a guard in the SQ and RQ completion handlers to ignore
calls generated by some providers when the QP is destroyed.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Some providers may wait while destroying adapter resources.
Since it is possible that the last reference is put on the
dto_tasklet, the actual destroy must be scheduled as a work item.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
The rq_cq_reap function is only called from the dto_tasklet. The
only resource shared with other threads is the sc_rq_dto_q. Move the
spin lock to protect only this list.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
