Users would like to control the behaviour of rdma_cm.
For example, old applications which don't set the
required RoCE gid type could be executed on RoCE V2
network types. In order to support this configuration,
we implement a configfs for rdma_cm.
In order to use the configfs, one needs to mount it and
mkdir <IB device name> inside rdma_cm directory.
The patch adds support for a single configuration file,
default_roce_mode. The mode can either be "IB/RoCE v1" or
"RoCE v2".
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
This adds an abstraction that allows ULPs to simply pass a completion
object and completion callback with each submitted WR and let the RDMA
core handle the nitty gritty details of how to handle completion
interrupts and poll the CQ.
In detail there is a new ib_cqe structure which just contains the
completion callback, and which can be used to get at the containing
object using container_of. It is pointed to by the WR and WC as an
alternative to the wr_id field, similar to how many ULPs already use
the field to store a pointer using casts.
A driver using the new completion callbacks allocates it's CQs using
the new ib_create_cq API, which in addition to the number of CQEs and
the completion vectors also takes a mode on how we poll for CQEs.
Three modes are available: direct for drivers that never take CQ
interrupts and just poll for them, softirq to poll from softirq context
using the to be renamed blk-iopoll infrastructure which takes care of
rearming and budgeting, or a workqueue for consumer who want to be
called from user context.
Thanks a lot to Sagi Grimberg who helped reviewing the API, wrote
the current version of the workqueue code because my two previous
attempts sucked too much and converted the iSER initiator to the new
API.
Signed-off-by: Christoph Hellwig <hch@lst.de>
RoCE GIDs are based on IP addresses configured on Ethernet net-devices
which relate to the RDMA (RoCE) device port.
Currently, each of the low-level drivers that support RoCE (ocrdma,
mlx4) manages its own RoCE port GID table. As there's nothing which is
essentially vendor specific, we generalize that, and enhance the RDMA
core GID cache to do this job.
In order to populate the GID table, we listen for events:
(a) netdev up/down/change_addr events - if a netdev is built onto
our RoCE device, we need to add/delete its IPs. This involves
adding all GIDs related to this ndev, add default GIDs, etc.
(b) inet events - add new GIDs (according to the IP addresses)
to the table.
For programming the port RoCE GID table, providers must implement
the add_gid and del_gid callbacks.
RoCE GID management requires us to state the associated net_device
alongside the GID. This information is necessary in order to manage
the GID table. For example, when a net_device is removed, its
associated GIDs need to be removed as well.
RoCE mandates generating a default GID for each port, based on the
related net-device's IPv6 link local. In contrast to the GID based on
the regular IPv6 link-local (as we generate GID per IP address),
the default GID is also available when the net device is down (in
order to support loopback).
Locking is done as follows:
The patch modify the GID table code both for new RoCE drivers
implementing the add_gid/del_gid callbacks and for current RoCE and
IB drivers that do not. The flows for updating the table are
different, so the locking requirements are too.
While updating RoCE GID table, protection against multiple writers is
achieved via mutex_lock(&table->lock). Since writing to a table
requires us to find an entry (possible a free entry) in the table and
then modify it, this mutex protects both the find_gid and write_gid
ensuring the atomicity of the action.
Each entry in the GID cache is protected by rwlock. In RoCE, writing
(usually results from netdev notifier) involves invoking the vendor's
add_gid and del_gid callbacks, which could sleep.
Therefore, an invalid flag is added for each entry. Updates for RoCE are
done via a workqueue, thus sleeping is permitted.
In IB, updates are done in write_lock_irq(&device->cache.lock), thus
write_gid isn't allowed to sleep and add_gid/del_gid are not called.
When passing net-device into/out-of the GID cache, the device
is always passed held (dev_hold).
The code uses a single work item for updating all RDMA devices,
following a netdev or inet notifier.
The patch moves the cache from being a client (which was incorrect,
as the cache is part of the IB infrastructure) to being explicitly
initialized/freed when a device is registered/removed.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
* Add an interval tree implementation for ODP umems. Create an
interval tree for each ucontext (including a count of the number of
ODP MRs in this context, semaphore, etc.), and register ODP umems in
the interval tree.
* Add MMU notifiers handling functions, using the interval tree to
notify only the relevant umems and underlying MRs.
* Register to receive MMU notifier events from the MM subsystem upon
ODP MR registration (and unregister accordingly).
* Add a completion object to synchronize the destruction of ODP umems.
* Add mechanism to abort page faults when there's a concurrent invalidation.
The way we synchronize between concurrent invalidations and page
faults is by keeping a counter of currently running invalidations, and
a sequence number that is incremented whenever an invalidation is
caught. The page fault code checks the counter and also verifies that
the sequence number hasn't progressed before it updates the umem's
page tables. This is similar to what the kvm module does.
In order to prevent the case where we register a umem in the middle of
an ongoing notifier, we also keep a per ucontext counter of the total
number of active mmu notifiers. We only enable new umems when all the
running notifiers complete.
Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: Shachar Raindel <raindel@mellanox.com>
Signed-off-by: Haggai Eran <haggaie@mellanox.com>
Signed-off-by: Yuval Dagan <yuvalda@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
* Extend the umem struct to keep the ODP related data.
* Allocate and initialize the ODP related information in the umem
(page_list, dma_list) and freeing as needed in the end of the run.
* Store a reference to the process PID struct in the ucontext. Used to
safely obtain the task_struct and the mm during fault handling,
without preventing the task destruction if needed.
* Add 2 helper functions: ib_umem_odp_map_dma_pages and
ib_umem_odp_unmap_dma_pages. These functions get the DMA addresses
of specific pages of the umem (and, currently, pin them).
* Support for page faults only - IB core will keep the reference on
the pages used and call put_page when freeing an ODP umem
area. Invalidations support will be added in a later patch.
Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: Shachar Raindel <raindel@mellanox.com>
Signed-off-by: Haggai Eran <haggaie@mellanox.com>
Signed-off-by: Majd Dibbiny <majd@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
This patch adds iWARP Port Mapper (IWPM) Version 2 support. The iWARP
Port Mapper implementation is based on the port mapper specification
section in the Sockets Direct Protocol paper -
http://www.rdmaconsortium.org/home/draft-pinkerton-iwarp-sdp-v1.0.pdf
Existing iWARP RDMA providers use the same IP address as the native
TCP/IP stack when creating RDMA connections. They need a mechanism to
claim the TCP ports used for RDMA connections to prevent TCP port
collisions when other host applications use TCP ports. The iWARP Port
Mapper provides a standard mechanism to accomplish this. Without this
service it is possible for RDMA application to bind/listen on the same
port which is already being used by native TCP host application. If
that happens the incoming TCP connection data can be passed to the
RDMA stack with error.
The iWARP Port Mapper solution doesn't contain any changes to the
existing network stack in the kernel space. All the changes are
contained with the infiniband tree and also in user space.
The iWARP Port Mapper service is implemented as a user space daemon
process. Source for the IWPM service is located at
http://git.openfabrics.org/git?p=~tnikolova/libiwpm-1.0.0/.git;a=summary
The iWARP driver (port mapper client) sends to the IWPM service the
local IP address and TCP port it has received from the RDMA
application, when starting a connection. The IWPM service performs a
socket bind from user space to get an available TCP port, called a
mapped port, and communicates it back to the client. In that sense,
the IWPM service is used to map the TCP port, which the RDMA
application uses to any port available from the host TCP port
space. The mapped ports are used in iWARP RDMA connections to avoid
collisions with native TCP stack which is aware that these ports are
taken. When an RDMA connection using a mapped port is terminated, the
client notifies the IWPM service, which then releases the TCP port.
The message exchange between the IWPM service and the iWARP drivers
(between user space and kernel space) is implemented using netlink
sockets.
1) Netlink interface functions are added: ibnl_unicast() and
ibnl_mulitcast() for sending netlink messages to user space
2) The signature of the existing ibnl_put_msg() is changed to be more
generic
3) Two netlink clients are added: RDMA_NL_NES, RDMA_NL_C4IW
corresponding to the two iWarp drivers - nes and cxgb4 which use
the IWPM service
4) Enums are added to enumerate the attributes in the netlink
messages, which are exchanged between the user space IWPM service
and the iWARP drivers
Signed-off-by: Tatyana Nikolova <tatyana.e.nikolova@intel.com>
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Reviewed-by: PJ Waskiewicz <pj.waskiewicz@solidfire.com>
[ Fold in range checking fixes and nlh_next removal as suggested by Dan
Carpenter and Steve Wise. Fix sparse endianness in hash. - Roland ]
Signed-off-by: Roland Dreier <roland@purestorage.com>
IP based addressing introduces the usage of rdma_addr_find_dmac_by_grh()
within ib_core. Since this function is declared in ib_addr, ib_addr
should be a part of the core INFINIBAND modules, rather than
INFINIBAND_ADDR_TRANS.
Signed-off-by: Matan Barak <matanb@mellanox.com>
Signed-off-by: Roland Dreier <roland@purestorage.com>
Add basic RDMA netlink infrastructure that allows for registration of
RDMA clients for which data is to be exported and supplies message
construction callbacks.
Signed-off-by: Nir Muchtar <nirm@voltaire.com>
[ Reorganize a few things, add CONFIG_NET dependency. - Roland ]
Signed-off-by: Roland Dreier <roland@purestorage.com>
Export ib_umem_get()/ib_umem_release() and put low-level drivers in
control of when to call ib_umem_get() to pin and DMA map userspace,
rather than always calling it in ib_uverbs_reg_mr() before calling the
low-level driver's reg_user_mr method.
Also move these functions to be in the ib_core module instead of
ib_uverbs, so that driver modules using them do not depend on
ib_uverbs.
This has a number of advantages:
- It is better design from the standpoint of making generic code a
library that can be used or overridden by device-specific code as
the details of specific devices dictate.
- Drivers that do not need to pin userspace memory regions do not
need to take the performance hit of calling ib_mem_get(). For
example, although I have not tried to implement it in this patch,
the ipath driver should be able to avoid pinning memory and just
use copy_{to,from}_user() to access userspace memory regions.
- Buffers that need special mapping treatment can be identified by
the low-level driver. For example, it may be possible to solve
some Altix-specific memory ordering issues with mthca CQs in
userspace by mapping CQ buffers with extra flags.
- Drivers that need to pin and DMA map userspace memory for things
other than memory regions can use ib_umem_get() directly, instead
of hacks using extra parameters to their reg_phys_mr method. For
example, the mlx4 driver that is pending being merged needs to pin
and DMA map QP and CQ buffers, but it does not need to create a
memory key for these buffers. So the cleanest solution is for mlx4
to call ib_umem_get() in the create_qp and create_cq methods.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
The IB SA tracks multicast join/leave requests on a per port basis and
does not do any reference counting: if two users of the same port join
the same group, and one leaves that group, then the SA will remove the
port from the group even though there is one user who wants to stay a
member left. Therefore, in order to support multiple users of the
same multicast group from the same port, we need to perform reference
counting locally.
To do this, add an multicast submodule to ib_sa to perform reference
counting of multicast join/leave operations. Modify ib_ipoib (the
only in-kernel user of multicast) to use the new interface.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Export the rdma cm interfaces to userspace via a misc device.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Modifications to the existing rdma header files, core files, drivers,
and ulp files to support iWARP, including:
- Hook iWARP CM into the build system and use it in rdma_cm.
- Convert enum ib_node_type to enum rdma_node_type, which includes
the possibility of RDMA_NODE_RNIC, and update everything for this.
Signed-off-by: Tom Tucker <tom@opengridcomputing.com>
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Kernel connection management agent over InfiniBand that connects based
on IP addresses. The agent defines a generic RDMA connection
abstraction to support clients wanting to connect over different RDMA
devices.
The agent also handles RDMA device hotplug events on behalf of clients.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add an address translation service that maps IP addresses to
InfiniBand GID addresses using IPoIB.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Provide common handling for marshalling data between userspace clients
and kernel InfiniBand drivers.
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add a new config option INFINIBAND_USER_MAD to control whether we
build ib_umad. Change INFINIBAND_USER_VERBS to INFINIBAND_USER_ACCESS,
and have it control ib_ucm and ib_uat as well as ib_uverbs.
Signed-off-by: James Lentini <jlentini@netapp.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Move the InfiniBand headers from drivers/infiniband/include to include/rdma.
This allows InfiniBand-using code to live elsewhere, and lets us remove the
ugly EXTRA_CFLAGS include path from the InfiniBand Makefiles.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Hook up userspace CM to the make system
Signed-off-by: Libor Michalek <libor@topspin.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Add the kernel CM implementation
Signed-off-by: Sean Hefty <sean.hefty@intel.com>
Signed-off-by: Hal Rosenstock <halr@voltaire.com>
Cc: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Hook up InfiniBand userspace verbs to Kconfig and the make system.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
