5113f6f70
 김현기
kernel add
|
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
|
Introduction
============
The IBM Power architecture provides support for CAPI (Coherent
Accelerator Power Interface), which is available to certain PCIe slots
on Power 8 systems. CAPI can be thought of as a special tunneling
protocol through PCIe that allow PCIe adapters to look like special
purpose co-processors which can read or write an application's
memory and generate page faults. As a result, the host interface to
an adapter running in CAPI mode does not require the data buffers to
be mapped to the device's memory (IOMMU bypass) nor does it require
memory to be pinned.
On Linux, Coherent Accelerator (CXL) kernel services present CAPI
devices as a PCI device by implementing a virtual PCI host bridge.
This abstraction simplifies the infrastructure and programming
model, allowing for drivers to look similar to other native PCI
device drivers.
CXL provides a mechanism by which user space applications can
directly talk to a device (network or storage) bypassing the typical
kernel/device driver stack. The CXL Flash Adapter Driver enables a
user space application direct access to Flash storage.
The CXL Flash Adapter Driver is a kernel module that sits in the
SCSI stack as a low level device driver (below the SCSI disk and
protocol drivers) for the IBM CXL Flash Adapter. This driver is
responsible for the initialization of the adapter, setting up the
special path for user space access, and performing error recovery. It
communicates directly the Flash Accelerator Functional Unit (AFU)
as described in Documentation/powerpc/cxl.txt.
The cxlflash driver supports two, mutually exclusive, modes of
operation at the device (LUN) level:
- Any flash device (LUN) can be configured to be accessed as a
regular disk device (i.e.: /dev/sdc). This is the default mode.
- Any flash device (LUN) can be configured to be accessed from
user space with a special block library. This mode further
specifies the means of accessing the device and provides for
either raw access to the entire LUN (referred to as direct
or physical LUN access) or access to a kernel/AFU-mediated
partition of the LUN (referred to as virtual LUN access). The
segmentation of a disk device into virtual LUNs is assisted
by special translation services provided by the Flash AFU.
Overview
========
The Coherent Accelerator Interface Architecture (CAIA) introduces a
concept of a master context. A master typically has special privileges
granted to it by the kernel or hypervisor allowing it to perform AFU
wide management and control. The master may or may not be involved
directly in each user I/O, but at the minimum is involved in the
initial setup before the user application is allowed to send requests
directly to the AFU.
The CXL Flash Adapter Driver establishes a master context with the
AFU. It uses memory mapped I/O (MMIO) for this control and setup. The
Adapter Problem Space Memory Map looks like this:
+-------------------------------+
| 512 * 64 KB User MMIO |
| (per context) |
| User Accessible |
+-------------------------------+
| 512 * 128 B per context |
| Provisioning and Control |
| Trusted Process accessible |
+-------------------------------+
| 64 KB Global |
| Trusted Process accessible |
+-------------------------------+
This driver configures itself into the SCSI software stack as an
adapter driver. The driver is the only entity that is considered a
Trusted Process to program the Provisioning and Control and Global
areas in the MMIO Space shown above. The master context driver
discovers all LUNs attached to the CXL Flash adapter and instantiates
scsi block devices (/dev/sdb, /dev/sdc etc.) for each unique LUN
seen from each path.
Once these scsi block devices are instantiated, an application
written to a specification provided by the block library may get
access to the Flash from user space (without requiring a system call).
This master context driver also provides a series of ioctls for this
block library to enable this user space access. The driver supports
two modes for accessing the block device.
The first mode is called a virtual mode. In this mode a single scsi
block device (/dev/sdb) may be carved up into any number of distinct
virtual LUNs. The virtual LUNs may be resized as long as the sum of
the sizes of all the virtual LUNs, along with the meta-data associated
with it does not exceed the physical capacity.
The second mode is called the physical mode. In this mode a single
block device (/dev/sdb) may be opened directly by the block library
and the entire space for the LUN is available to the application.
Only the physical mode provides persistence of the data. i.e. The
data written to the block device will survive application exit and
restart and also reboot. The virtual LUNs do not persist (i.e. do
not survive after the application terminates or the system reboots).
Block library API
=================
Applications intending to get access to the CXL Flash from user
space should use the block library, as it abstracts the details of
interfacing directly with the cxlflash driver that are necessary for
performing administrative actions (i.e.: setup, tear down, resize).
The block library can be thought of as a 'user' of services,
implemented as IOCTLs, that are provided by the cxlflash driver
specifically for devices (LUNs) operating in user space access
mode. While it is not a requirement that applications understand
the interface between the block library and the cxlflash driver,
a high-level overview of each supported service (IOCTL) is provided
below.
The block library can be found on GitHub:
http://www.github.com/mikehollinger/ibmcapikv
CXL Flash Driver IOCTLs
=======================
Users, such as the block library, that wish to interface with a flash
device (LUN) via user space access need to use the services provided
by the cxlflash driver. As these services are implemented as ioctls,
a file descriptor handle must first be obtained in order to establish
the communication channel between a user and the kernel. This file
descriptor is obtained by opening the device special file associated
with the scsi disk device (/dev/sdb) that was created during LUN
discovery. As per the location of the cxlflash driver within the
SCSI protocol stack, this open is actually not seen by the cxlflash
driver. Upon successful open, the user receives a file descriptor
(herein referred to as fd1) that should be used for issuing the
subsequent ioctls listed below.
The structure definitions for these IOCTLs are available in:
uapi/scsi/cxlflash_ioctl.h
DK_CXLFLASH_ATTACH
------------------
This ioctl obtains, initializes, and starts a context using the CXL
kernel services. These services specify a context id (u16) by which
to uniquely identify the context and its allocated resources. The
services additionally provide a second file descriptor (herein
referred to as fd2) that is used by the block library to initiate
memory mapped I/O (via mmap()) to the CXL flash device and poll for
completion events. This file descriptor is intentionally installed by
this driver and not the CXL kernel services to allow for intermediary
notification and access in the event of a non-user-initiated close(),
such as a killed process. This design point is described in further
detail in the description for the DK_CXLFLASH_DETACH ioctl.
There are a few important aspects regarding the "tokens" (context id
and fd2) that are provided back to the user:
- These tokens are only valid for the process under which they
were created. The child of a forked process cannot continue
to use the context id or file descriptor created by its parent
(see DK_CXLFLASH_VLUN_CLONE for further details).
- These tokens are only valid for the lifetime of the context and
the process under which they were created. Once either is
destroyed, the tokens are to be considered stale and subsequent
usage will result in errors.
- When a context is no longer needed, the user shall detach from
the context via the DK_CXLFLASH_DETACH ioctl.
- A close on fd2 will invalidate the tokens. This operation is not
required by the user.
DK_CXLFLASH_USER_DIRECT
-----------------------
This ioctl is responsible for transitioning the LUN to direct
(physical) mode access and configuring the AFU for direct access from
user space on a per-context basis. Additionally, the block size and
last logical block address (LBA) are returned to the user.
As mentioned previously, when operating in user space access mode,
LUNs may be accessed in whole or in part. Only one mode is allowed
at a time and if one mode is active (outstanding references exist),
requests to use the LUN in a different mode are denied.
The AFU is configured for direct access from user space by adding an
entry to the AFU's resource handle table. The index of the entry is
treated as a resource handle that is returned to the user. The user
is then able to use the handle to reference the LUN during I/O.
DK_CXLFLASH_USER_VIRTUAL
------------------------
This ioctl is responsible for transitioning the LUN to virtual mode
of access and configuring the AFU for virtual access from user space
on a per-context basis. Additionally, the block size and last logical
block address (LBA) are returned to the user.
As mentioned previously, when operating in user space access mode,
LUNs may be accessed in whole or in part. Only one mode is allowed
at a time and if one mode is active (outstanding references exist),
requests to use the LUN in a different mode are denied.
The AFU is configured for virtual access from user space by adding
an entry to the AFU's resource handle table. The index of the entry
is treated as a resource handle that is returned to the user. The
user is then able to use the handle to reference the LUN during I/O.
By default, the virtual LUN is created with a size of 0. The user
would need to use the DK_CXLFLASH_VLUN_RESIZE ioctl to adjust the grow
the virtual LUN to a desired size. To avoid having to perform this
resize for the initial creation of the virtual LUN, the user has the
option of specifying a size as part of the DK_CXLFLASH_USER_VIRTUAL
ioctl, such that when success is returned to the user, the
resource handle that is provided is already referencing provisioned
storage. This is reflected by the last LBA being a non-zero value.
DK_CXLFLASH_VLUN_RESIZE
-----------------------
This ioctl is responsible for resizing a previously created virtual
LUN and will fail if invoked upon a LUN that is not in virtual
mode. Upon success, an updated last LBA is returned to the user
indicating the new size of the virtual LUN associated with the
resource handle.
The partitioning of virtual LUNs is jointly mediated by the cxlflash
driver and the AFU. An allocation table is kept for each LUN that is
operating in the virtual mode and used to program a LUN translation
table that the AFU references when provided with a resource handle.
DK_CXLFLASH_RELEASE
-------------------
This ioctl is responsible for releasing a previously obtained
reference to either a physical or virtual LUN. This can be
thought of as the inverse of the DK_CXLFLASH_USER_DIRECT or
DK_CXLFLASH_USER_VIRTUAL ioctls. Upon success, the resource handle
is no longer valid and the entry in the resource handle table is
made available to be used again.
As part of the release process for virtual LUNs, the virtual LUN
is first resized to 0 to clear out and free the translation tables
associated with the virtual LUN reference.
DK_CXLFLASH_DETACH
------------------
This ioctl is responsible for unregistering a context with the
cxlflash driver and release outstanding resources that were
not explicitly released via the DK_CXLFLASH_RELEASE ioctl. Upon
success, all "tokens" which had been provided to the user from the
DK_CXLFLASH_ATTACH onward are no longer valid.
DK_CXLFLASH_VLUN_CLONE
----------------------
This ioctl is responsible for cloning a previously created
context to a more recently created context. It exists solely to
support maintaining user space access to storage after a process
forks. Upon success, the child process (which invoked the ioctl)
will have access to the same LUNs via the same resource handle(s)
and fd2 as the parent, but under a different context.
Context sharing across processes is not supported with CXL and
therefore each fork must be met with establishing a new context
for the child process. This ioctl simplifies the state management
and playback required by a user in such a scenario. When a process
forks, child process can clone the parents context by first creating
a context (via DK_CXLFLASH_ATTACH) and then using this ioctl to
perform the clone from the parent to the child.
The clone itself is fairly simple. The resource handle and lun
translation tables are copied from the parent context to the child's
and then synced with the AFU.
DK_CXLFLASH_VERIFY
------------------
This ioctl is used to detect various changes such as the capacity of
the disk changing, the number of LUNs visible changing, etc. In cases
where the changes affect the application (such as a LUN resize), the
cxlflash driver will report the changed state to the application.
The user calls in when they want to validate that a LUN hasn't been
changed in response to a check condition. As the user is operating out
of band from the kernel, they will see these types of events without
the kernel's knowledge. When encountered, the user's architected
behavior is to call in to this ioctl, indicating what they want to
verify and passing along any appropriate information. For now, only
verifying a LUN change (ie: size different) with sense data is
supported.
DK_CXLFLASH_RECOVER_AFU
-----------------------
This ioctl is used to drive recovery (if such an action is warranted)
of a specified user context. Any state associated with the user context
is re-established upon successful recovery.
User contexts are put into an error condition when the device needs to
be reset or is terminating. Users are notified of this error condition
by seeing all 0xF's on an MMIO read. Upon encountering this, the
architected behavior for a user is to call into this ioctl to recover
their context. A user may also call into this ioctl at any time to
check if the device is operating normally. If a failure is returned
from this ioctl, the user is expected to gracefully clean up their
context via release/detach ioctls. Until they do, the context they
hold is not relinquished. The user may also optionally exit the process
at which time the context/resources they held will be freed as part of
the release fop.
DK_CXLFLASH_MANAGE_LUN
----------------------
This ioctl is used to switch a LUN from a mode where it is available
for file-system access (legacy), to a mode where it is set aside for
exclusive user space access (superpipe). In case a LUN is visible
across multiple ports and adapters, this ioctl is used to uniquely
identify each LUN by its World Wide Node Name (WWNN).
|