Extra resources for handling backchannel requests have to be
pre-allocated when a transport instance is created. Set up
additional fields in svcxprt_rdma to track these resources.
The max_requests fields are elements of the RPC-over-RDMA
protocol, so they should be u32. To ensure that unsigned
arithmetic is used everywhere, some other fields in the
svcxprt_rdma struct are updated.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Acked-by: Bruce Fields <bfields@fieldses.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Pre-requisite to use map_xdr in the backchannel code.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Acked-by: Bruce Fields <bfields@fieldses.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Clean up.
These functions can otherwise fail, so check for page allocation
failures too.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Acked-by: Bruce Fields <bfields@fieldses.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
svc_rdma_post_recv() allocates pages for receive buffers on-demand.
It uses GFP_KERNEL so the allocator tries hard, and may sleep. But
I'm about to add a call to svc_rdma_post_recv() from a function
that may not sleep.
Since all svc_rdma_post_recv() call sites can tolerate its failure,
allow it to fail if the page allocator returns nothing. Longer term,
receive buffers, being a finite resource per-connection, should be
pre-allocated and re-used.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Acked-by: Bruce Fields <bfields@fieldses.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
To ensure this allocation cannot fail and will not sleep,
pre-allocate the req_map structures per-connection.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Acked-by: Bruce Fields <bfields@fieldses.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
When the maximum payload size of NFS READ and WRITE was increased
by commit cc9a903d91 ("svcrdma: Change maximum server payload back
to RPCSVC_MAXPAYLOAD"), the size of struct svc_rdma_op_ctxt
increased to over 6KB (on x86_64). That makes allocating one of
these from a kmem_cache more likely to fail in situations when
system memory is exhausted.
Since I'm about to add a caller where this allocation must always
work _and_ it cannot sleep, pre-allocate ctxts for each connection.
Another motivation for this change is that NFSv4.x servers are
required by specification not to drop NFS requests. Pre-allocating
memory resources reduces the likelihood of a drop.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Acked-by: Bruce Fields <bfields@fieldses.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Be sure the completed ctxt is put in every path.
The xprt enqueue can take a while, so put the completed ctxt back
in circulation _before_ enqueuing the xprt.
Remove/disable debugging.
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Acked-by: Bruce Fields <bfields@fieldses.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
kzalloc is used here, so setting the atomic fields to zero is
unnecessary. sc_ord is set again in handle_connect_req. The other
fields are re-initialized in svc_rdma_accept().
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Acked-by: Bruce Fields <bfields@fieldses.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Previously, IPV6_DEFAULT_HOPLIMIT was used as the hop limit value for
RoCE. Fixing that by taking ip4_dst_hoplimit and ip6_dst_hoplimit as
hop limit values.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
rdma_addr_find_dmac_by_grh resolves dmac, vlan_id and if_index and
downsteram patch will also add hop_limit as an output parameter,
thus we rename it to rdma_addr_find_l2_eth_by_grh.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
The IRQ_POLL_F_SCHED bit is set as long as polling is ongoing.
This means that irq_poll_sched() must proceed if this bit has
not yet been set.
Fixes: commit ea51190c03 ("irq_poll: fold irq_poll_sched_prep into irq_poll_sched").
Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Sparse complains about dereference before check. Fixing this by
moving the check before the dereference.
Fixes: 200298326b ('IB/core: Validate route when we init ah')
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
When write_gid function needs to do a sleep-able operation, it unlocks
table->rwlock and then relocks it. Sparse complains about context
imbalance.
This is safe as write_gid is always called with table->rwlock.
write_gid protects from simultaneous writes to this GID entry
by setting the GID_TABLE_ENTRY_INVALID flag.
Fixes: 9c584f0495 ('IB/core: Change per-entry lock in RoCE GID table to
one lock')
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Port number is not part of ClassPortInfo attribute but is
still needed as a parameter when invoking process_mad.
To properly handle this attribute, port_num is added as a
parameter to get_counter_table and get_perf_mad was changed
not to store port_num in the attribute itself when it's
querying the ClassPortInfo attribute.
This handles issue pointed out by Matan Barak <matanb@dev.mellanox.co.il>
Fixes: 145d9c5410 ('IB/core: Display extended counter set if available')
Signed-off-by: Hal Rosenstock <hal@mellanox.com>
Acked-by: Matan Barak <matanb@mellanox.com>
Acked-by: Ira Weiny <ira.weiny@intel.com>
Reviewed-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Detected this by building the IB core with W=1. See also patch
"IB core: Fix ib_sg_to_pages()" (commit 8f5ba10ed4).
Signed-off-by: Bart Van Assche <bart.vanassche@sandisk.com>
Cc: Sagi Grimberg <sagig@mellanox.com>
Cc: Christoph Hellwig <hch@lst.de>
Reviewed-by: Leon Romanovsky <leon.romanovsky@mellanox.com>
Acked-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Remove the local workqueue to process mad completions and use the CQ API
instead.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hal Rosenstock <hal@mellanox.com>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Stop abusing wr_id and just pass the parameter explicitly.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hal Rosenstock <hal@mellanox.com>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Use eth_zero_addr to assign the zero address to the given address
array instead of memset when second argument is address of zero.
Signed-off-by: Lucas Tanure <tanure@linux.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Fix the following sparse warning:
drivers/infiniband/hw/mlx5/main.c:1061:29: warning: symbol 'pfn' shadows
an earlier one
drivers/infiniband/hw/mlx5/main.c:1030:21: originally declared here
Fixes: d69e3bcf79 ('IB/mlx5: Mmap the HCA's core clock register to user-space')
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
The macro mlx4_foreach_non_ib_transport_port() is not used anywhere. Remove it.
Fixes: aa9a2d51a3 ("mlx4: Activate RoCE/SRIOV")
Signed-off-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
In commit dbf727de74 ("IB/core: Use GID table in AH creation and dmac
resolution") we copy source mac to mlx4_ah from the attributes of
gid at ib_ah_attr.grh.sgid_index. Now we can use it.
Signed-off-by: Moni Shoua <monis@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Hop limit value wasn't copied from attributes when ah was created.
This may influence packets for unconnected services to get dropped in
routers when endpoints are not in the same subnet.
Fixes: fa417f7b52 ("IB/mlx4: Add support for IBoE")
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Maximum number of EQE capacity per CQ was mistakenly exposed
as CQE. Fix that.
Fixes: 938fe83c8d ("net/mlx5_core: New device capabilities handling")
Signed-off-by: Leon Romanovsky <leonro@mellanox.com>
Cc: <stable@vger.kernel.org>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
The h/w is designed in such a way that, if you do anything IPv6
related, a valid clip entry must be there. So take clip reference
before creating IPv6 listening servers, and then if we fail to
create server, release the clip entry.
Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Acked-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Commit c5dfb000b9 ("iw_cxgb4: Pass qid range to user space driver")
from Dec 11, 2015, leads to the following static checker warning:
drivers/infiniband/hw/cxgb4/device.c:857 c4iw_rdev_open()
warn: variable dereferenced before check 'rdev->status_page'
Also we weren't deallocating ocqp pool in error path when failed to
allocate status page. Fixing it too.
Fixes: c5dfb000b9 ("iw_cxgb4: Pass qid range to user space driver")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Hariprasad Shenai <hariprasad@chelsio.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
The issue here is that there is a cut and paste bug. When we allocate
cma_dev_group->default_ports_group we use "sizeof(*cma_dev_group->ports)"
instead of "sizeof(*cma_dev_group->default_ports_group)".
We're bumping up against the 80 character limit so I introduced a new
local pointer "ports_group" to get around that.
Fixes: 045959db65 ('IB/cma: Add configfs for rdma_cm')
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
nes_reg_phys_mr() returns ERR_PTRs on error. It doesn't return NULL.
This bug has been there for a while, but we recently changed from
calling a function pointer to calling nes_reg_phys_mr() directly so now
Smatch is able to detect the bug.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Doug Ledford <dledford@redhat.com>
The current code is problematic when the QP creation and ipoib is used to
support NFS and NFS desires to do IO for paging purposes. In that case, the
GFP_KERNEL allocation in qib_qp.c causes a deadlock in tight memory
situations.
This fix adds support to create queue pair with GFP_NOIO flag for connected
mode only to cleanly fail the create queue pair in those situations.
Cc: <stable@vger.kernel.org> # 3.16+
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Vinit Agnihotri <vinit.abhay.agnihotri@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Allocate private data structure at probe and release at disconnect.
The private data will be used to track the connections managed by the
driver, but for now only holds a count of the number of cports the
bundle has.
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Use the gb_connection_mutex when making sure a remote CPort is
available.
This is needed when moving to driver-managed connections that can be
created and destroyed at any time.
Also check for duplicate bundle-less connections.
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Serialise connection creation against concurrent creation and
destruction using a global mutex.
This is needed to prevent two drivers from attempting to create a
connection to the same interface CPort and to cope with a racing
connection destroy when moving to driver managed connections.
Note that the locking can not (easily) be made more fine-grained as not
all connections have an interface, but these are not hot paths anyway.
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Fix longstanding race that could lead to a connection being destroyed
while processing an incoming message due to missing reference counting
when looking up the connection.
Add helpers to manipulate the connection kref.
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Drop the useless connection_mutex that did not, and can not be used to
prevent connection lookup races in atomic context and therefore serves
no purpose.
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Destroy the work queue when the connection is destroyed/deregistered
instead of when the last reference is dropped.
The work queue is not needed once the connection has been deregistered,
and no operations will ever be added to it again (handled by checking
connection->state) even if the connection object may not be deallocated
until the final reference is dropped.
The work-queue name is set based on the host-device name and host-device
cport id, something which guarantees a unique name. This would no longer
be true if the work queue was not destroyed at connection deregistration
as a new connection could then be created for that very same host cport.
This is not necessarily a problem unless some work queue features are
used that require unique work-queue names, but let's try to be well
behaved.
Also update an obsolete comment and make explicit that a connection must
be disabled before being destroyed.
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Remove WARN_ON from connection destroy testing for a NULL-connection
when destroying a connection.
This will allow for simpler driver code when releasing driver-managed
connections.
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Add bundle-version operation to fetch the version of the bundle class.
Retrieve the bundle version of all bundles when initialising the
interface in case the control-protocol version is greater than 0.1.
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Remove the broken legacy protocol-version handling from core and move it
to the legacy driver instead.
Note that version handling has always been broken as legacy protocols
were looked up and bound to connections before the connections were
enabled and version negotiation could take place (over that very
connection). Document that in the legacy driver as well.
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Store the "negotiated" protocol version to use in the svc state
struct instead of the connection struct.
The generic concept of a connection version is going away in favour of
bundle-class versions.
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Disable bundle connections in core before calling driver disconnect in
case the interface is already gone.
This avoids unnecessary timeouts on hot-unplug when a driver does I/O in
its disconnect callback.
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Use the protocol pointer to determine the legacy connection state.
This is needed to allow core disable connections when an interface has
been hot-unplugged without the legacy protocols leaking its resources.
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Move legacy protocol and connection handling to the legacy driver.
Rename the former global functions using a common legacy_ prefix.
Note that all legacy protocols suffer from a connection initialisation
race in that the protocol-specific initialisation, which includes
allocation of protocol-specific state containers, can not happen until
*after* the connection has been enabled. This is a major flaw in the
original design that we can now finally address by converting the legacy
protocol drivers into proper bundle (class) drivers.
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Drop dependency on the legacy protocol abstraction.
Remove the now unused and last legacy-protocol flag
GB_PROTOCOL_SKIP_VERSION along with the protocol-flag feature.
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Drop dependency on the legacy protocol abstraction.
Instead implement the protocol-specific version request directly, and
use the new interface for managing the control connection.
Note that the version request is being removed from most protocols, but
we need to keep the current request for the control protocol as-is
indefinitely to maintain backwards compatibility (e.g. with the ES2/ES3
bootrom).
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Remove control-protocol version from the exported protocol definitions
as it is an implementation detail that makes no sense to export.
Currently gbsim uses the kernel's control-protocol version definitions
directly instead of reporting the version of the protocol it actually
implements.
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
Remove the legacy protocol flags that were used to suppress the
connected and disconnected events.
Instead send the connected and disconnected event for all bundle
connections and explicitly exclude static connections and control
connections.
Reviewed-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Johan Hovold <johan@hovoldconsulting.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@google.com>
