-INFINIBAND MIDLAYER LOCKING
+===========================
+InfiniBand Midlayer Locking
+===========================
This guide is an attempt to make explicit the locking assumptions
made by the InfiniBand midlayer. It describes the requirements on
protocols that use the midlayer.
Sleeping and interrupt context
+==============================
With the following exceptions, a low-level driver implementation of
all of the methods in struct ib_device may sleep. The exceptions
are any methods from the list:
- create_ah
- modify_ah
- query_ah
- destroy_ah
- post_send
- post_recv
- poll_cq
- req_notify_cq
- map_phys_fmr
+ - create_ah
+ - modify_ah
+ - query_ah
+ - destroy_ah
+ - post_send
+ - post_recv
+ - poll_cq
+ - req_notify_cq
+ - map_phys_fmr
which may not sleep and must be callable from any context.
The corresponding functions exported to upper level protocol
consumers:
- ib_create_ah
- ib_modify_ah
- ib_query_ah
- ib_destroy_ah
- ib_post_send
- ib_post_recv
- ib_req_notify_cq
- ib_map_phys_fmr
+ - ib_create_ah
+ - ib_modify_ah
+ - ib_query_ah
+ - ib_destroy_ah
+ - ib_post_send
+ - ib_post_recv
+ - ib_req_notify_cq
+ - ib_map_phys_fmr
are therefore safe to call from any context.
In addition, the function
- ib_dispatch_event
+ - ib_dispatch_event
used by low-level drivers to dispatch asynchronous events through
the midlayer is also safe to call from any context.
Reentrancy
+----------
All of the methods in struct ib_device exported by a low-level
driver must be fully reentrant. The low-level driver is required to
information between different calls of ib_poll_cq() is not defined.
Callbacks
+---------
A low-level driver must not perform a callback directly from the
same callchain as an ib_device method call. For example, it is not
completion event handlers for the same CQ are not called
simultaneously. The driver must guarantee that only one CQ event
handler for a given CQ is running at a time. In other words, the
- following situation is not allowed:
+ following situation is not allowed::
- CPU1 CPU2
+ CPU1 CPU2
- low-level driver ->
- consumer CQ event callback:
- /* ... */
- ib_req_notify_cq(cq, ...);
- low-level driver ->
- /* ... */ consumer CQ event callback:
- /* ... */
- return from CQ event handler
+ low-level driver ->
+ consumer CQ event callback:
+ /* ... */
+ ib_req_notify_cq(cq, ...);
+ low-level driver ->
+ /* ... */ consumer CQ event callback:
+ /* ... */
+ return from CQ event handler
The context in which completion event and asynchronous event
callbacks run is not defined. Depending on the low-level driver, it
Upper level protocol consumers may not sleep in a callback.
Hot-plug
+--------
A low-level driver announces that a device is ready for use by
consumers when it calls ib_register_device(), all initialization
--- /dev/null
+:orphan:
+
+==========
+InfiniBand
+==========
+
+.. toctree::
+ :maxdepth: 1
+
+ core_locking
+ ipoib
+ opa_vnic
+ sysfs
+ tag_matching
+ user_mad
+ user_verbs
+
+.. only:: subproject and html
+
+ Indices
+ =======
+
+ * :ref:`genindex`
-IP OVER INFINIBAND
+==================
+IP over InfiniBand
+==================
The ib_ipoib driver is an implementation of the IP over InfiniBand
protocol as specified by RFC 4391 and 4392, issued by the IETF ipoib
masqueraded to the kernel as ethernet interfaces).
Partitions and P_Keys
+=====================
When the IPoIB driver is loaded, it creates one interface for each
port using the P_Key at index 0. To create an interface with a
different P_Key, write the desired P_Key into the main interface's
- /sys/class/net/<intf name>/create_child file. For example:
+ /sys/class/net/<intf name>/create_child file. For example::
echo 0x8001 > /sys/class/net/ib0/create_child
This will create an interface named ib0.8001 with P_Key 0x8001. To
- remove a subinterface, use the "delete_child" file:
+ remove a subinterface, use the "delete_child" file::
echo 0x8001 > /sys/class/net/ib0/delete_child
rtnl_link_ops, where children created using either way behave the same.
Datagram vs Connected modes
+===========================
The IPoIB driver supports two modes of operation: datagram and
connected. The mode is set and read through an interface's
networking stack to use the smaller UD MTU for these neighbours.
Stateless offloads
+==================
If the IB HW supports IPoIB stateless offloads, IPoIB advertises
TCP/IP checksum and/or Large Send (LSO) offloading capability to the
on/off using ethtool calls. Currently LRO is supported only for
checksum offload capable devices.
- Stateless offloads are supported only in datagram mode.
+ Stateless offloads are supported only in datagram mode.
Interrupt moderation
+====================
If the underlying IB device supports CQ event moderation, one can
use ethtool to set interrupt mitigation parameters and thus reduce
moderation is supported.
Debugging Information
+=====================
By compiling the IPoIB driver with CONFIG_INFINIBAND_IPOIB_DEBUG set
to 'y', tracing messages are compiled into the driver. They are
runtime through files in /sys/module/ib_ipoib/.
CONFIG_INFINIBAND_IPOIB_DEBUG also enables files in the debugfs
- virtual filesystem. By mounting this filesystem, for example with
+ virtual filesystem. By mounting this filesystem, for example with::
mount -t debugfs none /sys/kernel/debug
performance, because it adds tests to the fast path.
References
+==========
Transmission of IP over InfiniBand (IPoIB) (RFC 4391)
- http://ietf.org/rfc/rfc4391.txt
+ http://ietf.org/rfc/rfc4391.txt
+
IP over InfiniBand (IPoIB) Architecture (RFC 4392)
- http://ietf.org/rfc/rfc4392.txt
+ http://ietf.org/rfc/rfc4392.txt
+
IP over InfiniBand: Connected Mode (RFC 4755)
http://ietf.org/rfc/rfc4755.txt
+=================================================================
+Intel Omni-Path (OPA) Virtual Network Interface Controller (VNIC)
+=================================================================
+
Intel Omni-Path (OPA) Virtual Network Interface Controller (VNIC) feature
supports Ethernet functionality over Omni-Path fabric by encapsulating
the Ethernet packets between HFI nodes.
Ethernet Manager (EM) which is part of the trusted Fabric Manager (FM)
application. HFI nodes can have multiple VNICs each connected to a
different virtual Ethernet switch. The below diagram presents a case
-of two virtual Ethernet switches with two HFI nodes.
-
- +-------------------+
- | Subnet/ |
- | Ethernet |
- | Manager |
- +-------------------+
- / /
- / /
- / /
- / /
-+-----------------------------+ +------------------------------+
-| Virtual Ethernet Switch | | Virtual Ethernet Switch |
-| +---------+ +---------+ | | +---------+ +---------+ |
-| | VPORT | | VPORT | | | | VPORT | | VPORT | |
-+--+---------+----+---------+-+ +-+---------+----+---------+---+
- | \ / |
- | \ / |
- | \/ |
- | / \ |
- | / \ |
- +-----------+------------+ +-----------+------------+
- | VNIC | VNIC | | VNIC | VNIC |
- +-----------+------------+ +-----------+------------+
- | HFI | | HFI |
- +------------------------+ +------------------------+
+of two virtual Ethernet switches with two HFI nodes::
+
+ +-------------------+
+ | Subnet/ |
+ | Ethernet |
+ | Manager |
+ +-------------------+
+ / /
+ / /
+ / /
+ / /
+ +-----------------------------+ +------------------------------+
+ | Virtual Ethernet Switch | | Virtual Ethernet Switch |
+ | +---------+ +---------+ | | +---------+ +---------+ |
+ | | VPORT | | VPORT | | | | VPORT | | VPORT | |
+ +--+---------+----+---------+-+ +-+---------+----+---------+---+
+ | \ / |
+ | \ / |
+ | \/ |
+ | / \ |
+ | / \ |
+ +-----------+------------+ +-----------+------------+
+ | VNIC | VNIC | | VNIC | VNIC |
+ +-----------+------------+ +-----------+------------+
+ | HFI | | HFI |
+ +------------------------+ +------------------------+
The Omni-Path encapsulated Ethernet packet format is as described below.
-Bits Field
-------------------------------------
+==================== ================================
+Bits Field
+==================== ================================
Quad Word 0:
-0-19 SLID (lower 20 bits)
-20-30 Length (in Quad Words)
-31 BECN bit
-32-51 DLID (lower 20 bits)
-52-56 SC (Service Class)
-57-59 RC (Routing Control)
-60 FECN bit
-61-62 L2 (=10, 16B format)
-63 LT (=1, Link Transfer Head Flit)
+0-19 SLID (lower 20 bits)
+20-30 Length (in Quad Words)
+31 BECN bit
+32-51 DLID (lower 20 bits)
+52-56 SC (Service Class)
+57-59 RC (Routing Control)
+60 FECN bit
+61-62 L2 (=10, 16B format)
+63 LT (=1, Link Transfer Head Flit)
Quad Word 1:
-0-7 L4 type (=0x78 ETHERNET)
-8-11 SLID[23:20]
-12-15 DLID[23:20]
-16-31 PKEY
-32-47 Entropy
-48-63 Reserved
+0-7 L4 type (=0x78 ETHERNET)
+8-11 SLID[23:20]
+12-15 DLID[23:20]
+16-31 PKEY
+32-47 Entropy
+48-63 Reserved
Quad Word 2:
-0-15 Reserved
-16-31 L4 header
-32-63 Ethernet Packet
+0-15 Reserved
+16-31 L4 header
+32-63 Ethernet Packet
Quad Words 3 to N-1:
-0-63 Ethernet packet (pad extended)
+0-63 Ethernet packet (pad extended)
Quad Word N (last):
-0-23 Ethernet packet (pad extended)
-24-55 ICRC
-56-61 Tail
-62-63 LT (=01, Link Transfer Tail Flit)
+0-23 Ethernet packet (pad extended)
+24-55 ICRC
+56-61 Tail
+62-63 LT (=01, Link Transfer Tail Flit)
+==================== ================================
Ethernet packet is padded on the transmit side to ensure that the VNIC OPA
packet is quad word aligned. The 'Tail' field contains the number of bytes
Omni-Path header in the transmit path. For each VNIC interface, the
information required for encapsulation is configured by the EM via VEMA MAD
interface. It also passes any control information to the HW dependent driver
-by invoking the RDMA netdev control operations.
+by invoking the RDMA netdev control operations::
+-------------------+ +----------------------+
| | | Linux |
-SYSFS FILES
+===========
+Sysfs files
+===========
The sysfs interface has moved to
Documentation/ABI/stable/sysfs-class-infiniband.
+==================
Tag matching logic
+==================
The MPI standard defines a set of rules, known as tag-matching, for matching
source send operations to destination receives. The following parameters must
match the following source and destination parameters:
+
* Communicator
* User tag - wild card may be specified by the receiver
* Source rank – wild car may be specified by the receiver
* Destination rank – wild
+
The ordering rules require that when more than one pair of send and receive
message envelopes may match, the pair that includes the earliest posted-send
and the earliest posted-receive is the pair that must be used to satisfy the
A fin message needs to be received in order for the buffer to be reused.
Tag matching implementation
+===========================
There are two types of matching objects used, the posted receive list and the
unexpected message list. The application posts receive buffers through calls
-USERSPACE MAD ACCESS
+====================
+Userspace MAD access
+====================
Device files
+============
Each port of each InfiniBand device has a "umad" device and an
"issm" device attached. For example, a two-port HCA will have two
device of each type (for switch port 0).
Creating MAD agents
+===================
A MAD agent can be created by filling in a struct ib_user_mad_reg_req
and then calling the IB_USER_MAD_REGISTER_AGENT ioctl on a file
descriptor for the appropriate device file. If the registration
request succeeds, a 32-bit id will be returned in the structure.
- For example:
+ For example::
struct ib_user_mad_reg_req req = { /* ... */ };
ret = ioctl(fd, IB_USER_MAD_REGISTER_AGENT, (char *) &req);
ioctl. Also, all agents registered through a file descriptor will
be unregistered when the descriptor is closed.
- 2014 -- a new registration ioctl is now provided which allows additional
+ 2014
+ a new registration ioctl is now provided which allows additional
fields to be provided during registration.
Users of this registration call are implicitly setting the use of
pkey_index (see below).
Receiving MADs
+==============
MADs are received using read(). The receive side now supports
RMPP. The buffer passed to read() must be at least one
MAD (RMPP), the errno is set to ENOSPC and the length of the
buffer needed is set in mad.length.
- Example for normal MAD (non RMPP) reads:
+ Example for normal MAD (non RMPP) reads::
+
struct ib_user_mad *mad;
mad = malloc(sizeof *mad + 256);
ret = read(fd, mad, sizeof *mad + 256);
free(mad);
}
- Example for RMPP reads:
+ Example for RMPP reads::
+
struct ib_user_mad *mad;
mad = malloc(sizeof *mad + 256);
ret = read(fd, mad, sizeof *mad + 256);
poll()/select() may be used to wait until a MAD can be read.
Sending MADs
+============
MADs are sent using write(). The agent ID for sending should be
filled into the id field of the MAD, the destination LID should be
filled into the lid field, and so on. The send side does support
- RMPP so arbitrary length MAD can be sent. For example:
+ RMPP so arbitrary length MAD can be sent. For example::
struct ib_user_mad *mad;
perror("write");
Transaction IDs
+===============
Users of the umad devices can use the lower 32 bits of the
transaction ID field (that is, the least significant half of the
the kernel and will be overwritten before a MAD is sent.
P_Key Index Handling
+====================
The old ib_umad interface did not allow setting the P_Key index for
MADs that are sent and did not provide a way for obtaining the P_Key
default, and the IB_USER_MAD_ENABLE_PKEY ioctl will be removed.
Setting IsSM Capability Bit
+===========================
To set the IsSM capability bit for a port, simply open the
corresponding issm device file. If the IsSM bit is already set,
the issm file.
/dev files
+==========
To create the appropriate character device files automatically with
- udev, a rule like
+ udev, a rule like::
KERNEL=="umad*", NAME="infiniband/%k"
KERNEL=="issm*", NAME="infiniband/%k"
- can be used. This will create device nodes named
+ can be used. This will create device nodes named::
/dev/infiniband/umad0
/dev/infiniband/issm0
for the first port, and so on. The InfiniBand device and port
- associated with these devices can be determined from the files
+ associated with these devices can be determined from the files::
/sys/class/infiniband_mad/umad0/ibdev
/sys/class/infiniband_mad/umad0/port
- and
+ and::
/sys/class/infiniband_mad/issm0/ibdev
/sys/class/infiniband_mad/issm0/port
-USERSPACE VERBS ACCESS
+======================
+Userspace verbs access
+======================
The ib_uverbs module, built by enabling CONFIG_INFINIBAND_USER_VERBS,
enables direct userspace access to IB hardware via "verbs," as
libmthca userspace driver be installed.
User-kernel communication
+=========================
Userspace communicates with the kernel for slow path, resource
management operations via the /dev/infiniband/uverbsN character
system call.
Resource management
+===================
Since creation and destruction of all IB resources is done by
commands passed through a file descriptor, the kernel can keep track
prevent one process from touching another process's resources.
Memory pinning
+==============
Direct userspace I/O requires that memory regions that are potential
I/O targets be kept resident at the same physical address. The
number of pages pinned by a process.
/dev files
+==========
To create the appropriate character device files automatically with
- udev, a rule like
+ udev, a rule like::
KERNEL=="uverbs*", NAME="infiniband/%k"
- can be used. This will create device nodes named
+ can be used. This will create device nodes named::
/dev/infiniband/uverbs0
"process %s did not enable P_Key index support.\n",
current->comm);
dev_warn(&file->port->dev,
- " Documentation/infiniband/user_mad.txt has info on the new ABI.\n");
+ " Documentation/infiniband/user_mad.rst has info on the new ABI.\n");
}
}
transports IP packets over InfiniBand so you can use your IB
device as a fancy NIC.
- See Documentation/infiniband/ipoib.txt for more information
+ See Documentation/infiniband/ipoib.rst for more information
config INFINIBAND_IPOIB_CM
bool "IP-over-InfiniBand Connected Mode support"