summaryrefslogtreecommitdiff
path: root/docs/pvrdma.txt
blob: 5c122fe81813e793e395d834d7b1c2321086102f (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
Paravirtualized RDMA Device (PVRDMA)
====================================


1. Description
===============
PVRDMA is the QEMU implementation of VMware's paravirtualized RDMA device.
It works with its Linux Kernel driver AS IS, no need for any special guest
modifications.

While it complies with the VMware device, it can also communicate with bare
metal RDMA-enabled machines as peers.

It does not require an RDMA HCA in the host, it can work with Soft-RoCE (rxe).

It does not require the whole guest RAM to be pinned allowing memory
over-commit and, even if not implemented yet, migration support will be
possible with some HW assistance.

A project presentation accompany this document:
- https://blog.linuxplumbersconf.org/2017/ocw/system/presentations/4730/original/lpc-2017-pvrdma-marcel-apfelbaum-yuval-shaia.pdf



2. Setup
========


2.1 Guest setup
===============
Fedora 27+ kernels work out of the box, older distributions
require updating the kernel to 4.14 to include the pvrdma driver.

However the libpvrdma library needed by User Level Software is still
not available as part of the distributions, so the rdma-core library
needs to be compiled and optionally installed.

Please follow the instructions at:
  https://github.com/linux-rdma/rdma-core.git


2.2 Host Setup
==============
The pvrdma backend is an ibdevice interface that can be exposed
either by a Soft-RoCE(rxe) device on machines with no RDMA device,
or an HCA SRIOV function(VF/PF).
Note that ibdevice interfaces can't be shared between pvrdma devices,
each one requiring a separate instance (rxe or SRIOV VF).


2.2.1 Soft-RoCE backend(rxe)
===========================
A stable version of rxe is required, Fedora 27+ or a Linux
Kernel 4.14+ is preferred.

The rdma_rxe module is part of the Linux Kernel but not loaded by default.
Install the User Level library (librxe) following the instructions from:
https://github.com/SoftRoCE/rxe-dev/wiki/rxe-dev:-Home

Associate an ETH interface with rxe by running:
   rxe_cfg add eth0
An rxe0 ibdevice interface will be created and can be used as pvrdma backend.


2.2.2 RDMA device Virtual Function backend
==========================================
Nothing special is required, the pvrdma device can work not only with
Ethernet Links, but also Infinibands Links.
All is needed is an ibdevice with an active port, for Mellanox cards
will be something like mlx5_6 which can be the backend.


2.2.3 QEMU setup
================
Configure QEMU with --enable-rdma flag, installing
the required RDMA libraries.



3. Usage
========


3.1 VM Memory settings
======================
Currently the device is working only with memory backed RAM
and it must be mark as "shared":
   -m 1G \
   -object memory-backend-ram,id=mb1,size=1G,share \
   -numa node,memdev=mb1 \


3.2 MAD Multiplexer
===================
MAD Multiplexer is a service that exposes MAD-like interface for VMs in
order to overcome the limitation where only single entity can register with
MAD layer to send and receive RDMA-CM MAD packets.

To build rdmacm-mux run
# make rdmacm-mux

Before running the rdmacm-mux make sure that both ib_cm and rdma_cm kernel
modules aren't loaded, otherwise the rdmacm-mux service will fail to start.

The application accepts 3 command line arguments and exposes a UNIX socket
to pass control and data to it.
-d rdma-device-name  Name of RDMA device to register with
-s unix-socket-path  Path to unix socket to listen (default /var/run/rdmacm-mux)
-p rdma-device-port  Port number of RDMA device to register with (default 1)
The final UNIX socket file name is a concatenation of the 3 arguments so
for example for device mlx5_0 on port 2 this /var/run/rdmacm-mux-mlx5_0-2
will be created.

pvrdma requires this service.

Please refer to contrib/rdmacm-mux for more details.


3.3 Service exposed by libvirt daemon
=====================================
The control over the RDMA device's GID table is done by updating the
device's Ethernet function addresses.
Usually the first GID entry is determined by the MAC address, the second by
the first IPv6 address and the third by the IPv4 address. Other entries can
be added by adding more IP addresses. The opposite is the same, i.e.
whenever an address is removed, the corresponding GID entry is removed.
The process is done by the network and RDMA stacks. Whenever an address is
added the ib_core driver is notified and calls the device driver add_gid
function which in turn update the device.
To support this in pvrdma device the device hooks into the create_bind and
destroy_bind HW commands triggered by pvrdma driver in guest.

Whenever changed is made to the pvrdma port's GID table a special QMP
messages is sent to be processed by libvirt to update the address of the
backend Ethernet device.

pvrdma requires that libvirt service will be up.


3.4 PCI devices settings
========================
RoCE device exposes two functions - an Ethernet and RDMA.
To support it, pvrdma device is composed of two PCI functions, an Ethernet
device of type vmxnet3 on PCI slot 0 and a PVRDMA device on PCI slot 1. The
Ethernet function can be used for other Ethernet purposes such as IP.


3.5 Device parameters
=====================
- netdev: Specifies the Ethernet device function name on the host for
  example enp175s0f0. For Soft-RoCE device (rxe) this would be the Ethernet
  device used to create it.
- ibdev: The IB device name on host for example rxe0, mlx5_0 etc.
- mad-chardev: The name of the MAD multiplexer char device.
- ibport: In case of multi-port device (such as Mellanox's HCA) this
  specify the port to use. If not set 1 will be used.
- dev-caps-max-mr-size: The maximum size of MR.
- dev-caps-max-qp:      Maximum number of QPs.
- dev-caps-max-cq:      Maximum number of CQs.
- dev-caps-max-mr:      Maximum number of MRs.
- dev-caps-max-pd:      Maximum number of PDs.
- dev-caps-max-ah:      Maximum number of AHs.

Notes:
- The first 3 parameters are mandatory settings, the rest have their
  defaults.
- The last 8 parameters (the ones that prefixed by dev-caps) defines the top
  limits but the final values is adjusted by the backend device limitations.
- netdev can be extracted from ibdev's sysfs
  (/sys/class/infiniband/<ibdev>/device/net/)


3.6 Example
===========
Define bridge device with vmxnet3 network backend:
<interface type='bridge'>
  <mac address='56:b4:44:e9:62:dc'/>
  <source bridge='bridge1'/>
  <model type='vmxnet3'/>
  <address type='pci' domain='0x0000' bus='0x00' slot='0x10' function='0x0' multifunction='on'/>
</interface>

Define pvrdma device:
<qemu:commandline>
  <qemu:arg value='-object'/>
  <qemu:arg value='memory-backend-ram,id=mb1,size=1G,share'/>
  <qemu:arg value='-numa'/>
  <qemu:arg value='node,memdev=mb1'/>
  <qemu:arg value='-chardev'/>
  <qemu:arg value='socket,path=/var/run/rdmacm-mux-rxe0-1,id=mads'/>
  <qemu:arg value='-device'/>
  <qemu:arg value='pvrdma,addr=10.1,ibdev=rxe0,netdev=bridge0,mad-chardev=mads'/>
</qemu:commandline>



4. Implementation details
=========================


4.1 Overview
============
The device acts like a proxy between the Guest Driver and the host
ibdevice interface.
On configuration path:
 - For every hardware resource request (PD/QP/CQ/...) the pvrdma will request
   a resource from the backend interface, maintaining a 1-1 mapping
   between the guest and host.
On data path:
 - Every post_send/receive received from the guest will be converted into
   a post_send/receive for the backend. The buffers data will not be touched
   or copied resulting in near bare-metal performance for large enough buffers.
 - Completions from the backend interface will result in completions for
   the pvrdma device.


4.2 PCI BARs
============
PCI Bars:
	BAR 0 - MSI-X
        MSI-X vectors:
		(0) Command - used when execution of a command is completed.
		(1) Async - not in use.
		(2) Completion - used when a completion event is placed in
		  device's CQ ring.
	BAR 1 - Registers
        --------------------------------------------------------
        | VERSION |  DSR | CTL | REQ | ERR |  ICR | IMR  | MAC |
        --------------------------------------------------------
		DSR - Address of driver/device shared memory used
              for the command channel, used for passing:
			    - General info such as driver version
			    - Address of 'command' and 'response'
			    - Address of async ring
			    - Address of device's CQ ring
			    - Device capabilities
		CTL - Device control operations (activate, reset etc)
		IMG - Set interrupt mask
		REQ - Command execution register
		ERR - Operation status

	BAR 2 - UAR
        ---------------------------------------------------------
        | QP_NUM  | SEND/RECV Flag ||  CQ_NUM |   ARM/POLL Flag |
        ---------------------------------------------------------
		- Offset 0 used for QP operations (send and recv)
		- Offset 4 used for CQ operations (arm and poll)


4.3 Major flows
===============

4.3.1 Create CQ
===============
    - Guest driver
        - Allocates pages for CQ ring
        - Creates page directory (pdir) to hold CQ ring's pages
        - Initializes CQ ring
        - Initializes 'Create CQ' command object (cqe, pdir etc)
        - Copies the command to 'command' address
        - Writes 0 into REQ register
    - Device
        - Reads the request object from the 'command' address
        - Allocates CQ object and initialize CQ ring based on pdir
        - Creates the backend CQ
        - Writes operation status to ERR register
        - Posts command-interrupt to guest
    - Guest driver
        - Reads the HW response code from ERR register

4.3.2 Create QP
===============
    - Guest driver
        - Allocates pages for send and receive rings
        - Creates page directory(pdir) to hold the ring's pages
        - Initializes 'Create QP' command object (max_send_wr,
          send_cq_handle, recv_cq_handle, pdir etc)
        - Copies the object to 'command' address
        - Write 0 into REQ register
    - Device
        - Reads the request object from 'command' address
        - Allocates the QP object and initialize
            - Send and recv rings based on pdir
            - Send and recv ring state
        - Creates the backend QP
        - Writes the operation status to ERR register
        - Posts command-interrupt to guest
    - Guest driver
        - Reads the HW response code from ERR register

4.3.3 Post receive
==================
    - Guest driver
        - Initializes a wqe and place it on recv ring
        - Write to qpn|qp_recv_bit (31) to QP offset in UAR
    - Device
        - Extracts qpn from UAR
        - Walks through the ring and does the following for each wqe
            - Prepares the backend CQE context to be used when
              receiving completion from backend (wr_id, op_code, emu_cq_num)
            - For each sge prepares backend sge
            - Calls backend's post_recv

4.3.4 Process backend events
============================
    - Done by a dedicated thread used to process backend events;
      at initialization is attached to the device and creates
      the communication channel.
    - Thread main loop:
        - Polls for completions
        - Extracts QEMU _cq_num, wr_id and op_code from context
        - Writes CQE to CQ ring
        - Writes CQ number to device CQ
        - Sends completion-interrupt to guest
        - Deallocates context
        - Acks the event to backend



5. Limitations
==============
- The device obviously is limited by the Guest Linux Driver features implementation
  of the VMware device API.
- Memory registration mechanism requires mremap for every page in the buffer in order
  to map it to a contiguous virtual address range. Since this is not the data path
  it should not matter much. If the default max mr size is increased, be aware that
  memory registration can take up to 0.5 seconds for 1GB of memory.
- The device requires target page size to be the same as the host page size,
  otherwise it will fail to init.
- QEMU cannot map guest RAM from a file descriptor if a pvrdma device is attached,
  so it can't work with huge pages. The limitation will be addressed in the future,
  however QEMU allocates Guest RAM with MADV_HUGEPAGE so if there are enough huge
  pages available, QEMU will use them. QEMU will fail to init if the requirements
  are not met.



6. Performance
==============
By design the pvrdma device exits on each post-send/receive, so for small buffers
the performance is affected; however for medium buffers it will became close to
bare metal and from 1MB buffers and  up it reaches bare metal performance.
(tested with 2 VMs, the pvrdma devices connected to 2 VFs of the same device)

All the above assumes no memory registration is done on data path.