1 .. SPDX-License-Identifier: GPL-2.0
4 ===================================================
5 Virtual Accelerator Switchboard (VAS) userspace API
6 ===================================================
11 Power9 processor introduced Virtual Accelerator Switchboard (VAS) which
12 allows both userspace and kernel communicate to co-processor
13 (hardware accelerator) referred to as the Nest Accelerator (NX). The NX
14 unit comprises of one or more hardware engines or co-processor types
15 such as 842 compression, GZIP compression and encryption. On power9,
16 userspace applications will have access to only GZIP Compression engine
17 which supports ZLIB and GZIP compression algorithms in the hardware.
19 To communicate with NX, kernel has to establish a channel or window and
20 then requests can be submitted directly without kernel involvement.
21 Requests to the GZIP engine must be formatted as a co-processor Request
22 Block (CRB) and these CRBs must be submitted to the NX using COPY/PASTE
23 instructions to paste the CRB to hardware address that is associated with
24 the engine's request queue.
26 The GZIP engine provides two priority levels of requests: Normal and
27 High. Only Normal requests are supported from userspace right now.
29 This document explains userspace API that is used to interact with
30 kernel to setup channel / window which can be used to send compression
31 requests directly to NX accelerator.
37 Application access to the GZIP engine is provided through
38 /dev/crypto/nx-gzip device node implemented by the VAS/NX device driver.
39 An application must open the /dev/crypto/nx-gzip device to obtain a file
40 descriptor (fd). Then should issue VAS_TX_WIN_OPEN ioctl with this fd to
41 establish connection to the engine. It means send window is opened on GZIP
42 engine for this process. Once a connection is established, the application
43 should use the mmap() system call to map the hardware address of engine's
44 request queue into the application's virtual address space.
46 The application can then submit one or more requests to the engine by
47 using copy/paste instructions and pasting the CRBs to the virtual address
48 (aka paste_address) returned by mmap(). User space can close the
49 established connection or send window by closing the file descriptior
50 (close(fd)) or upon the process exit.
52 Note that applications can send several requests with the same window or
53 can establish multiple windows, but one window for each file descriptor.
55 Following sections provide additional details and references about the
61 There is one /dev/crypto/nx-gzip node in the system and it provides
62 access to all GZIP engines in the system. The only valid operations on
63 /dev/crypto/nx-gzip are:
65 * open() the device for read and write.
66 * issue VAS_TX_WIN_OPEN ioctl
67 * mmap() the engine's request queue into application's virtual
68 address space (i.e. get a paste_address for the co-processor
70 * close the device node.
72 Other file operations on this device node are undefined.
74 Note that the copy and paste operations go directly to the hardware and
75 do not go through this device. Refer COPY/PASTE document for more
78 Although a system may have several instances of the NX co-processor
79 engines (typically, one per P9 chip) there is just one
80 /dev/crypto/nx-gzip device node in the system. When the nx-gzip device
81 node is opened, Kernel opens send window on a suitable instance of NX
82 accelerator. It finds CPU on which the user process is executing and
83 determine the NX instance for the corresponding chip on which this CPU
86 Applications may chose a specific instance of the NX co-processor using
87 the vas_id field in the VAS_TX_WIN_OPEN ioctl as detailed below.
89 A userspace library libnxz is available here but still in development:
91 https://github.com/abalib/power-gzip
93 Applications that use inflate / deflate calls can link with libnxz
94 instead of libz and use NX GZIP compression without any modification.
96 Open /dev/crypto/nx-gzip
97 ========================
99 The nx-gzip device should be opened for read and write. No special
100 privileges are needed to open the device. Each window corresponds to one
101 file descriptor. So if the userspace process needs multiple windows,
102 several open calls have to be issued.
104 See open(2) system call man pages for other details such as return values,
105 error codes and restrictions.
107 VAS_TX_WIN_OPEN ioctl
108 =====================
110 Applications should use the VAS_TX_WIN_OPEN ioctl as follows to establish
111 a connection with NX co-processor engine:
115 struct vas_tx_win_open_attr {
117 __s16 vas_id; /* specific instance of vas or -1
120 __u64 flags; /* For future use */
125 The version field must be currently set to 1.
127 If '-1' is passed, kernel will make a best-effort attempt
128 to assign an optimal instance of NX for the process. To
129 select the specific VAS instance, refer
130 "Discovery of available VAS engines" section below.
132 flags, reserved1 and reserved2[6] fields are for future extension
133 and must be set to 0.
135 The attributes attr for the VAS_TX_WIN_OPEN ioctl are defined as
138 #define VAS_MAGIC 'v'
139 #define VAS_TX_WIN_OPEN _IOW(VAS_MAGIC, 1,
140 struct vas_tx_win_open_attr)
142 struct vas_tx_win_open_attr attr;
143 rc = ioctl(fd, VAS_TX_WIN_OPEN, &attr);
145 The VAS_TX_WIN_OPEN ioctl returns 0 on success. On errors, it
146 returns -1 and sets the errno variable to indicate the error.
150 ====== ================================================
151 EINVAL fd does not refer to a valid VAS device.
152 EINVAL Invalid vas ID
153 EINVAL version is not set with proper value
154 EEXIST Window is already opened for the given fd
155 ENOMEM Memory is not available to allocate window
156 ENOSPC System has too many active windows (connections)
158 EINVAL reserved fields are not set to 0.
159 ====== ================================================
161 See the ioctl(2) man page for more details, error codes and
164 mmap() NX-GZIP device
165 =====================
167 The mmap() system call for a NX-GZIP device fd returns a paste_address
168 that the application can use to copy/paste its CRB to the hardware engines.
172 paste_addr = mmap(addr, size, prot, flags, fd, offset);
174 Only restrictions on mmap for a NX-GZIP device fd are:
176 * size should be PAGE_SIZE
177 * offset parameter should be 0ULL
179 Refer to mmap(2) man page for additional details/restrictions.
180 In addition to the error conditions listed on the mmap(2) man
181 page, can also fail with one of the following error codes:
183 ====== =============================================
184 EINVAL fd is not associated with an open window
185 (i.e mmap() does not follow a successful call
186 to the VAS_TX_WIN_OPEN ioctl).
187 EINVAL offset field is not 0ULL.
188 ====== =============================================
190 Discovery of available VAS engines
191 ==================================
193 Each available VAS instance in the system will have a device tree node
194 like /proc/device-tree/vas@* or /proc/device-tree/xscom@*/vas@*.
195 Determine the chip or VAS instance and use the corresponding ibm,vas-id
196 property value in this node to select specific VAS instance.
198 Copy/Paste operations
199 =====================
201 Applications should use the copy and paste instructions to send CRB to NX.
202 Refer section 4.4 in PowerISA for Copy/Paste instructions:
203 https://openpowerfoundation.org/?resource_lib=power-isa-version-3-0
205 CRB Specification and use NX
206 ============================
208 Applications should format requests to the co-processor using the
209 co-processor Request Block (CRBs). Refer NX-GZIP user's manual for the format
210 of CRB and use NX from userspace such as sending requests and checking
216 Applications send requests to NX and wait for the status by polling on
217 co-processor Status Block (CSB) flags. NX updates status in CSB after each
218 request is processed. Refer NX-GZIP user's manual for the format of CSB and
221 In case if NX encounters translation error (called NX page fault) on CSB
222 address or any request buffer, raises an interrupt on the CPU to handle the
223 fault. Page fault can happen if an application passes invalid addresses or
224 request buffers are not in memory. The operating system handles the fault by
225 updating CSB with the following data::
228 csb.cc = CSB_CC_FAULT_ADDRESS;
229 csb.ce = CSB_CE_TERMINATION;
230 csb.address = fault_address;
232 When an application receives translation error, it can touch or access
233 the page that has a fault address so that this page will be in memory. Then
234 the application can resend this request to NX.
236 If the OS can not update CSB due to invalid CSB address, sends SEGV signal
237 to the process who opened the send window on which the original request was
238 issued. This signal returns with the following siginfo struct::
240 siginfo.si_signo = SIGSEGV;
241 siginfo.si_errno = EFAULT;
242 siginfo.si_code = SEGV_MAPERR;
243 siginfo.si_addr = CSB adress;
245 In the case of multi-thread applications, NX send windows can be shared
246 across all threads. For example, a child thread can open a send window,
247 but other threads can send requests to NX using this window. These
248 requests will be successful even in the case of OS handling faults as long
249 as CSB address is valid. If the NX request contains an invalid CSB address,
250 the signal will be sent to the child thread that opened the window. But if
251 the thread is exited without closing the window and the request is issued
252 using this window. the signal will be issued to the thread group leader
253 (tgid). It is up to the application whether to ignore or handle these
256 NX-GZIP User's Manual:
257 https://github.com/libnxz/power-gzip/blob/master/power_nx_gzip_um.pdf
268 struct vas_setup_attr txattr;
270 fd = open("/dev/crypto/nx-gzip", O_RDWR);
272 fprintf(stderr, "open nx-gzip failed\n");
275 memset(&txattr, 0, sizeof(txattr));
278 rc = ioctl(fd, VAS_TX_WIN_OPEN,
279 (unsigned long)&txattr);
281 fprintf(stderr, "ioctl() n %d, error %d\n",
285 addr = mmap(NULL, 4096, PROT_READ|PROT_WRITE,
286 MAP_SHARED, fd, 0ULL);
287 if (addr == MAP_FAILED) {
288 fprintf(stderr, "mmap() failed, errno %d\n",
293 //Format CRB request with compression or
295 // Refer tests for vas_copy/vas_paste
296 vas_copy((&crb, 0, 1);
297 vas_paste(addr, 0, 1);
298 // Poll on csb.flags with timeout
299 // csb address is listed in CRB
301 close(fd) or window can be closed upon process exit
304 Refer https://github.com/abalib/power-gzip for tests or more