8K message sizes are pretty important usecase for RDS current
workloads so we make provison to have 8K mrs available from the pool.
Based on number of SG's in the RDS message, we pick a pool to use.
Also to make sure that we don't under utlise mrs when say 8k messages
are dominating which could lead to 8k pull being exhausted, we fall-back
to 1m pool till 8k pool recovers for use.
This helps to at least push ~55 kB/s bidirectional data which
is a nice improvement.
Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
All HCA drivers seems to popullate max_mr caps and few of
them do both max_mr and max_fmr.
Hence update RDS code to make use of max_mr.
Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
- Create drivers/staging/rdma
- Move amso1100 driver to staging/rdma and schedule for deletion
- Move ipath driver to staging/rdma and schedule for deletion
- Add hfi1 driver to staging/rdma and set TODO for move to regular tree
- Initial support for namespaces to be used on RDMA devices
- Add RoCE GID table handling to the RDMA core caching code
- Infrastructure to support handling of devices with differing
read and write scatter gather capabilities
- Various iSER updates
- Kill off unsafe usage of global mr registrations
- Update SRP driver
- Misc. mlx4 driver updates
- Support for the mr_alloc verb
- Support for a netlink interface between kernel and user space cache
daemon to speed path record queries and route resolution
- Ininitial support for safe hot removal of verbs devices
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1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=EHLg
-----END PGP SIGNATURE-----
Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma
Pull inifiniband/rdma updates from Doug Ledford:
"This is a fairly sizeable set of changes. I've put them through a
decent amount of testing prior to sending the pull request due to
that.
There are still a few fixups that I know are coming, but I wanted to
go ahead and get the big, sizable chunk into your hands sooner rather
than waiting for those last few fixups.
Of note is the fact that this creates what is intended to be a
temporary area in the drivers/staging tree specifically for some
cleanups and additions that are coming for the RDMA stack. We
deprecated two drivers (ipath and amso1100) and are waiting to hear
back if we can deprecate another one (ehca). We also put Intel's new
hfi1 driver into this area because it needs to be refactored and a
transfer library created out of the factored out code, and then it and
the qib driver and the soft-roce driver should all be modified to use
that library.
I expect drivers/staging/rdma to be around for three or four kernel
releases and then to go away as all of the work is completed and final
deletions of deprecated drivers are done.
Summary of changes for 4.3:
- Create drivers/staging/rdma
- Move amso1100 driver to staging/rdma and schedule for deletion
- Move ipath driver to staging/rdma and schedule for deletion
- Add hfi1 driver to staging/rdma and set TODO for move to regular
tree
- Initial support for namespaces to be used on RDMA devices
- Add RoCE GID table handling to the RDMA core caching code
- Infrastructure to support handling of devices with differing read
and write scatter gather capabilities
- Various iSER updates
- Kill off unsafe usage of global mr registrations
- Update SRP driver
- Misc mlx4 driver updates
- Support for the mr_alloc verb
- Support for a netlink interface between kernel and user space cache
daemon to speed path record queries and route resolution
- Ininitial support for safe hot removal of verbs devices"
* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma: (136 commits)
IB/ipoib: Suppress warning for send only join failures
IB/ipoib: Clean up send-only multicast joins
IB/srp: Fix possible protection fault
IB/core: Move SM class defines from ib_mad.h to ib_smi.h
IB/core: Remove unnecessary defines from ib_mad.h
IB/hfi1: Add PSM2 user space header to header_install
IB/hfi1: Add CSRs for CONFIG_SDMA_VERBOSITY
mlx5: Fix incorrect wc pkey_index assignment for GSI messages
IB/mlx5: avoid destroying a NULL mr in reg_user_mr error flow
IB/uverbs: reject invalid or unknown opcodes
IB/cxgb4: Fix if statement in pick_local_ip6adddrs
IB/sa: Fix rdma netlink message flags
IB/ucma: HW Device hot-removal support
IB/mlx4_ib: Disassociate support
IB/uverbs: Enable device removal when there are active user space applications
IB/uverbs: Explicitly pass ib_dev to uverbs commands
IB/uverbs: Fix race between ib_uverbs_open and remove_one
IB/uverbs: Fix reference counting usage of event files
IB/core: Make ib_dealloc_pd return void
IB/srp: Create an insecure all physical rkey only if needed
...
The pd now has a local_dma_lkey member which completely replaces
ib_get_dma_mr, use it instead.
Signed-off-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
An ib_client callback that is called with the lists_rwsem locked only for
read is protected from changes to the IB client lists, but not from
ib_unregister_device() freeing its client data. This is because
ib_unregister_device() will remove the device from the device list with
lists_rwsem locked for write, but perform the rest of the cleanup,
including the call to remove() without that lock.
Mark client data that is undergoing de-registration with a new going_down
flag in the client data context. Lock the client data list with lists_rwsem
for write in addition to using the spinlock, so that functions calling the
callback would be able to lock only lists_rwsem for read and let callbacks
sleep.
Since ib_unregister_client() now marks the client data context, no need for
remove() to search the context again, so pass the client data directly to
remove() callbacks.
Reviewed-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
Signed-off-by: Haggai Eran <haggaie@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
RDS FMR flush operation and also it races with connect/reconect
which happes a lot with RDS. FMR flush being on common rds_wq aggrevates
the problem. Lets push RDS FMR pool flush work to its own worker.
Signed-off-by: Santosh Shilimkar <ssantosh@kernel.org>
Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Open the sockets calling sock_create_kern() with the correct struct net
pointer, and use that struct net pointer when verifying the
address passed to rds_bind().
Signed-off-by: Sowmini Varadhan <sowmini.varadhan@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
With calls to modular infrastructure, these files really
needs the full module.h header. Call it out so some of the
cleanups of implicit and unrequired includes elsewhere can be
cleaned up.
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
The RDMA CM currently infers the QP type from the port space selected
by the user. In the future (eg with RDMA_PS_IB or XRC), there may not
be a 1-1 correspondence between port space and QP type. For netlink
export of RDMA CM state, we want to export the QP type to userspace,
so it is cleaner to explicitly associate a QP type to an ID.
Modify rdma_create_id() to allow the user to specify the QP type, and
use it to make our selections of datagram versus connected mode.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
With cmwq, there's no reason to use dedicated rds_ib_fmr_wq - it's not
in the memory reclaim path and the maximum number of concurrent work
items is bound by the number of devices. Drop it and use system_wq
instead. This rds_ib_fmr_init/exit() noops. Both removed.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Andy Grover <andy.grover@oracle.com>
The RDS protocol has lots of functions that should be
declared static. rds_message_get/add_version_extension is
removed since it defined but never used.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The RDS IB device list wasn't protected by any locking. Traversal in
both the get_mr and FMR flushing paths could race with additon and
removal.
List manipulation is done with RCU primatives and is protected by the
write side of a rwsem. The list traversal in the get_mr fast path is
protected by a rcu read critical section. The FMR list traversal is
more problematic because it can block while traversing the list. We
protect this with the read side of the rwsem.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
This patch moves the FMR flushing work in to its own mult-threaded work queue.
This is to maintain performance in preparation for returning the main krdsd
work queue back to a single threaded work queue to avoid deep-rooted
concurrency bugs.
This is also good because it further separates FMRs, which might be removed
some day, from the rest of the code base.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
IB connections were not being destroyed during rmmod.
First, recently IB device removal callback was changed to disconnect
connections that used the removing device rather than destroying them. So
connections with devices during rmmod were not being destroyed.
Second, rds_ib_destroy_nodev_conns() was being called before connections are
disassociated with devices. It would almost never find connections in the
nodev list.
We first get rid of rds_ib_destroy_conns(), which is no longer called, and
refactor the existing caller into the main body of the function and get rid of
the list and lock wrappers.
Then we call rds_ib_destroy_nodev_conns() *after* ib_unregister_client() has
removed the IB device from all the conns and put the conns on the nodev list.
The result is that IB connections are destroyed by rmmod.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
The RDS IB client removal callback can queue work to drop the final reference
to an IB device. We have to make sure that this function has returned before
we complete rmmod or the work threads can try to execute freed code.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
Currently IB device removal destroys connections which are associated with the
device. This prevents connections from being re-established when replacement
devices are added.
Instead we'll queue shutdown work on the connections as their devices are
removed. When we see that devices are added we triger connection attempts on
all connections that don't currently have a device.
The result is that RDS sockets can resume device-independent work (bcopy, not
RDMA) across IB device removal and restoration.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
The RDS IB client .remove callback used to free the rds_ibdev for the given
device unconditionally. This could race other users of the struct. This patch
adds refcounting so that we only free the rds_ibdev once all of its users are
done.
Many rds_ibdev users are tied to connections. We give the connection a
reference and change these users to reference the device in the connection
instead of looking it up in the IB client data. The only user of the IB client
data remaining is the first lookup of the device as connections are built up.
Incrementing the reference count of a device found in the IB client data could
race with final freeing so we use an RCU grace period to make sure that freeing
won't happen until those lookups are done.
MRs need the rds_ibdev to get at the pool that they're freed in to. They exist
outside a connection and many MRs can reference different devices from one
socket, so it was natural to have each MR hold a reference. MR refs can be
dropped from interrupt handlers and final device teardown can block so we push
it off to a work struct. Pool teardown had to be fixed to cancel its pending
work instead of deadlocking waiting for all queued work, including itself, to
finish.
MRs get their reference from the global device list, which gets a reference.
It is left unprotected by locks and remains racy. A simple global lock would
be a significant bottleneck. More scalable (complicated) locking should be
done carefully in a later patch.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
Allocate send/recv rings in memory that is node-local to the HCA.
This significantly helps performance.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
rds_ib_get_device is called very often as we turn an
ip address into a corresponding device structure. It currently
take a global spinlock as it walks different lists to find active
devices.
This commit changes the lists over to RCU, which isn't very complex
because they are not updated very often at all.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
We now ask the transport to give us a rm for the congestion
map, and then we handle it normally. Previously, the
transport defined a function that we would call to send
a congestion map.
Convert TCP and loop transports to new cong map method.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Instead of using a constant for initiator_depth and
responder_resources, read the per-QP values when the
device is enumerated, and then use these values when creating
the connection.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Implement a CMSG-based interface to do FADD and CSWP ops.
Alter send routines to handle atomic ops.
Add atomic counters to stats.
Add xmit_atomic() to struct rds_transport
Inline rds_ib_send_unmap_rdma into unmap_rm
Signed-off-by: Andy Grover <andy.grover@oracle.com>
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
The RDMA CM is intended to support the use of a loopback address
when establishing a connection; however, the behavior of the CM
when loopback addresses are used is confusing and does not always
work, depending on whether loopback was specified by the server,
the client, or both.
The defined behavior of rdma_bind_addr is to associate an RDMA
device with an rdma_cm_id, as long as the user specified a non-
zero address. (ie they weren't just trying to reserve a port)
Currently, if the loopback address is passed to rdam_bind_addr,
no device is associated with the rdma_cm_id. Fix this.
If a loopback address is specified by the client as the destination
address for a connection, it will fail to establish a connection.
This is true even if the server is listing across all addresses or
on the loopback address itself. The issue is that the server tries
to translate the IP address carried in the REQ message to a local
net_device address, which fails. The translation is not needed in
this case, since the REQ carries the actual HW address that should
be used.
Finally, cleanup loopback support to be more transport neutral.
Replace separate calls to get/set the sgid and dgid from the
device address to a single call that behaves correctly depending
on the format of the device address. And support both IPv4 and
IPv6 address formats.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
[ Fixed RDS build by s/ib_addr_get/rdma_addr_get/ - Roland ]
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Now that transports can be loaded in arbitrary order,
it is important for rds_trans_get_preferred() to look
for them in a particular order, instead of walking the list
until it finds a transport that works for a given address.
Now, each transport registers for a specific transport slot,
and these are ordered so that preferred transports come first,
and then if they are not loaded, other transports are queried.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
While FMRs allow significant flexibility in what size of pages they can use,
we really just want FMR pages to match CPU page size. Roland says we can
count on this always being supported, so this simplifies things.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This will be default cause IB connections to failover faster,
but allow a longer retry count to be used if desired.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
rdma_create_id() doesn't return NULL, only ERR_PTR().
Found by smatch (http://repo.or.cz/w/smatch.git). Compile tested.
regards,
dan carpenter
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This fixes a bug where a connection was unexpectedly
not on *any* list while being destroyed. It also
cleans up some code duplication and regularizes some
function names.
* Grab appropriate lock in conn_free() and explain in comment
* Ensure via locking that a conn is never not on either
a dev's list or the nodev list
* Add rds_xx_remove_conn() to match rds_xx_add_conn()
* Make rds_xx_add_conn() return void
* Rename remove_{,nodev_}conns() to
destroy_{,nodev_}conns() and unify their implementation
in a helper function
* Document lock ordering as nodev conn_lock before
dev_conn_lock
Reported-by: Yosef Etigin <yosefe@voltaire.com>
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Registers as an RDS transport and an IB client, and uses IB CM
API to allocate ids, queue pairs, and the rest of that fun stuff.
Signed-off-by: Andy Grover <andy.grover@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Santosh Shilimkar