Loading...
1/*
2 * Copyright (c) 2010, Microsoft Corporation.
3 *
4 * This program is free software; you can redistribute it and/or modify it
5 * under the terms and conditions of the GNU General Public License,
6 * version 2, as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope it will be useful, but WITHOUT
9 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
10 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
11 * more details.
12 *
13 * You should have received a copy of the GNU General Public License along with
14 * this program; if not, write to the Free Software Foundation, Inc., 59 Temple
15 * Place - Suite 330, Boston, MA 02111-1307 USA.
16 *
17 * Authors:
18 * Haiyang Zhang <haiyangz@microsoft.com>
19 * Hank Janssen <hjanssen@microsoft.com>
20 */
21#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
22
23#include <linux/kernel.h>
24#include <linux/init.h>
25#include <linux/module.h>
26#include <linux/slab.h>
27#include <linux/sysctl.h>
28#include <linux/reboot.h>
29#include <linux/hyperv.h>
30
31#include "hyperv_vmbus.h"
32
33#define SD_MAJOR 3
34#define SD_MINOR 0
35#define SD_VERSION (SD_MAJOR << 16 | SD_MINOR)
36
37#define SD_WS2008_MAJOR 1
38#define SD_WS2008_VERSION (SD_WS2008_MAJOR << 16 | SD_MINOR)
39
40#define TS_MAJOR 3
41#define TS_MINOR 0
42#define TS_VERSION (TS_MAJOR << 16 | TS_MINOR)
43
44#define TS_WS2008_MAJOR 1
45#define TS_WS2008_VERSION (TS_WS2008_MAJOR << 16 | TS_MINOR)
46
47#define HB_MAJOR 3
48#define HB_MINOR 0
49#define HB_VERSION (HB_MAJOR << 16 | HB_MINOR)
50
51#define HB_WS2008_MAJOR 1
52#define HB_WS2008_VERSION (HB_WS2008_MAJOR << 16 | HB_MINOR)
53
54static int sd_srv_version;
55static int ts_srv_version;
56static int hb_srv_version;
57static int util_fw_version;
58
59static void shutdown_onchannelcallback(void *context);
60static struct hv_util_service util_shutdown = {
61 .util_cb = shutdown_onchannelcallback,
62};
63
64static void timesync_onchannelcallback(void *context);
65static struct hv_util_service util_timesynch = {
66 .util_cb = timesync_onchannelcallback,
67};
68
69static void heartbeat_onchannelcallback(void *context);
70static struct hv_util_service util_heartbeat = {
71 .util_cb = heartbeat_onchannelcallback,
72};
73
74static struct hv_util_service util_kvp = {
75 .util_cb = hv_kvp_onchannelcallback,
76 .util_init = hv_kvp_init,
77 .util_deinit = hv_kvp_deinit,
78};
79
80static struct hv_util_service util_vss = {
81 .util_cb = hv_vss_onchannelcallback,
82 .util_init = hv_vss_init,
83 .util_deinit = hv_vss_deinit,
84};
85
86static struct hv_util_service util_fcopy = {
87 .util_cb = hv_fcopy_onchannelcallback,
88 .util_init = hv_fcopy_init,
89 .util_deinit = hv_fcopy_deinit,
90};
91
92static void perform_shutdown(struct work_struct *dummy)
93{
94 orderly_poweroff(true);
95}
96
97/*
98 * Perform the shutdown operation in a thread context.
99 */
100static DECLARE_WORK(shutdown_work, perform_shutdown);
101
102static void shutdown_onchannelcallback(void *context)
103{
104 struct vmbus_channel *channel = context;
105 u32 recvlen;
106 u64 requestid;
107 bool execute_shutdown = false;
108 u8 *shut_txf_buf = util_shutdown.recv_buffer;
109
110 struct shutdown_msg_data *shutdown_msg;
111
112 struct icmsg_hdr *icmsghdrp;
113 struct icmsg_negotiate *negop = NULL;
114
115 vmbus_recvpacket(channel, shut_txf_buf,
116 PAGE_SIZE, &recvlen, &requestid);
117
118 if (recvlen > 0) {
119 icmsghdrp = (struct icmsg_hdr *)&shut_txf_buf[
120 sizeof(struct vmbuspipe_hdr)];
121
122 if (icmsghdrp->icmsgtype == ICMSGTYPE_NEGOTIATE) {
123 vmbus_prep_negotiate_resp(icmsghdrp, negop,
124 shut_txf_buf, util_fw_version,
125 sd_srv_version);
126 } else {
127 shutdown_msg =
128 (struct shutdown_msg_data *)&shut_txf_buf[
129 sizeof(struct vmbuspipe_hdr) +
130 sizeof(struct icmsg_hdr)];
131
132 switch (shutdown_msg->flags) {
133 case 0:
134 case 1:
135 icmsghdrp->status = HV_S_OK;
136 execute_shutdown = true;
137
138 pr_info("Shutdown request received -"
139 " graceful shutdown initiated\n");
140 break;
141 default:
142 icmsghdrp->status = HV_E_FAIL;
143 execute_shutdown = false;
144
145 pr_info("Shutdown request received -"
146 " Invalid request\n");
147 break;
148 }
149 }
150
151 icmsghdrp->icflags = ICMSGHDRFLAG_TRANSACTION
152 | ICMSGHDRFLAG_RESPONSE;
153
154 vmbus_sendpacket(channel, shut_txf_buf,
155 recvlen, requestid,
156 VM_PKT_DATA_INBAND, 0);
157 }
158
159 if (execute_shutdown == true)
160 schedule_work(&shutdown_work);
161}
162
163/*
164 * Set guest time to host UTC time.
165 */
166static inline void do_adj_guesttime(u64 hosttime)
167{
168 s64 host_tns;
169 struct timespec host_ts;
170
171 host_tns = (hosttime - WLTIMEDELTA) * 100;
172 host_ts = ns_to_timespec(host_tns);
173
174 do_settimeofday(&host_ts);
175}
176
177/*
178 * Set the host time in a process context.
179 */
180
181struct adj_time_work {
182 struct work_struct work;
183 u64 host_time;
184};
185
186static void hv_set_host_time(struct work_struct *work)
187{
188 struct adj_time_work *wrk;
189
190 wrk = container_of(work, struct adj_time_work, work);
191 do_adj_guesttime(wrk->host_time);
192 kfree(wrk);
193}
194
195/*
196 * Synchronize time with host after reboot, restore, etc.
197 *
198 * ICTIMESYNCFLAG_SYNC flag bit indicates reboot, restore events of the VM.
199 * After reboot the flag ICTIMESYNCFLAG_SYNC is included in the first time
200 * message after the timesync channel is opened. Since the hv_utils module is
201 * loaded after hv_vmbus, the first message is usually missed. The other
202 * thing is, systime is automatically set to emulated hardware clock which may
203 * not be UTC time or in the same time zone. So, to override these effects, we
204 * use the first 50 time samples for initial system time setting.
205 */
206static inline void adj_guesttime(u64 hosttime, u8 flags)
207{
208 struct adj_time_work *wrk;
209 static s32 scnt = 50;
210
211 wrk = kmalloc(sizeof(struct adj_time_work), GFP_ATOMIC);
212 if (wrk == NULL)
213 return;
214
215 wrk->host_time = hosttime;
216 if ((flags & ICTIMESYNCFLAG_SYNC) != 0) {
217 INIT_WORK(&wrk->work, hv_set_host_time);
218 schedule_work(&wrk->work);
219 return;
220 }
221
222 if ((flags & ICTIMESYNCFLAG_SAMPLE) != 0 && scnt > 0) {
223 scnt--;
224 INIT_WORK(&wrk->work, hv_set_host_time);
225 schedule_work(&wrk->work);
226 } else
227 kfree(wrk);
228}
229
230/*
231 * Time Sync Channel message handler.
232 */
233static void timesync_onchannelcallback(void *context)
234{
235 struct vmbus_channel *channel = context;
236 u32 recvlen;
237 u64 requestid;
238 struct icmsg_hdr *icmsghdrp;
239 struct ictimesync_data *timedatap;
240 u8 *time_txf_buf = util_timesynch.recv_buffer;
241 struct icmsg_negotiate *negop = NULL;
242
243 vmbus_recvpacket(channel, time_txf_buf,
244 PAGE_SIZE, &recvlen, &requestid);
245
246 if (recvlen > 0) {
247 icmsghdrp = (struct icmsg_hdr *)&time_txf_buf[
248 sizeof(struct vmbuspipe_hdr)];
249
250 if (icmsghdrp->icmsgtype == ICMSGTYPE_NEGOTIATE) {
251 vmbus_prep_negotiate_resp(icmsghdrp, negop,
252 time_txf_buf,
253 util_fw_version,
254 ts_srv_version);
255 } else {
256 timedatap = (struct ictimesync_data *)&time_txf_buf[
257 sizeof(struct vmbuspipe_hdr) +
258 sizeof(struct icmsg_hdr)];
259 adj_guesttime(timedatap->parenttime, timedatap->flags);
260 }
261
262 icmsghdrp->icflags = ICMSGHDRFLAG_TRANSACTION
263 | ICMSGHDRFLAG_RESPONSE;
264
265 vmbus_sendpacket(channel, time_txf_buf,
266 recvlen, requestid,
267 VM_PKT_DATA_INBAND, 0);
268 }
269}
270
271/*
272 * Heartbeat functionality.
273 * Every two seconds, Hyper-V send us a heartbeat request message.
274 * we respond to this message, and Hyper-V knows we are alive.
275 */
276static void heartbeat_onchannelcallback(void *context)
277{
278 struct vmbus_channel *channel = context;
279 u32 recvlen;
280 u64 requestid;
281 struct icmsg_hdr *icmsghdrp;
282 struct heartbeat_msg_data *heartbeat_msg;
283 u8 *hbeat_txf_buf = util_heartbeat.recv_buffer;
284 struct icmsg_negotiate *negop = NULL;
285
286 vmbus_recvpacket(channel, hbeat_txf_buf,
287 PAGE_SIZE, &recvlen, &requestid);
288
289 if (recvlen > 0) {
290 icmsghdrp = (struct icmsg_hdr *)&hbeat_txf_buf[
291 sizeof(struct vmbuspipe_hdr)];
292
293 if (icmsghdrp->icmsgtype == ICMSGTYPE_NEGOTIATE) {
294 vmbus_prep_negotiate_resp(icmsghdrp, negop,
295 hbeat_txf_buf, util_fw_version,
296 hb_srv_version);
297 } else {
298 heartbeat_msg =
299 (struct heartbeat_msg_data *)&hbeat_txf_buf[
300 sizeof(struct vmbuspipe_hdr) +
301 sizeof(struct icmsg_hdr)];
302
303 heartbeat_msg->seq_num += 1;
304 }
305
306 icmsghdrp->icflags = ICMSGHDRFLAG_TRANSACTION
307 | ICMSGHDRFLAG_RESPONSE;
308
309 vmbus_sendpacket(channel, hbeat_txf_buf,
310 recvlen, requestid,
311 VM_PKT_DATA_INBAND, 0);
312 }
313}
314
315static int util_probe(struct hv_device *dev,
316 const struct hv_vmbus_device_id *dev_id)
317{
318 struct hv_util_service *srv =
319 (struct hv_util_service *)dev_id->driver_data;
320 int ret;
321
322 srv->recv_buffer = kmalloc(PAGE_SIZE * 4, GFP_KERNEL);
323 if (!srv->recv_buffer)
324 return -ENOMEM;
325 srv->channel = dev->channel;
326 if (srv->util_init) {
327 ret = srv->util_init(srv);
328 if (ret) {
329 ret = -ENODEV;
330 goto error1;
331 }
332 }
333
334 /*
335 * The set of services managed by the util driver are not performance
336 * critical and do not need batched reading. Furthermore, some services
337 * such as KVP can only handle one message from the host at a time.
338 * Turn off batched reading for all util drivers before we open the
339 * channel.
340 */
341
342 set_channel_read_state(dev->channel, false);
343
344 hv_set_drvdata(dev, srv);
345
346 /*
347 * Based on the host; initialize the framework and
348 * service version numbers we will negotiate.
349 */
350 switch (vmbus_proto_version) {
351 case (VERSION_WS2008):
352 util_fw_version = UTIL_WS2K8_FW_VERSION;
353 sd_srv_version = SD_WS2008_VERSION;
354 ts_srv_version = TS_WS2008_VERSION;
355 hb_srv_version = HB_WS2008_VERSION;
356 break;
357
358 default:
359 util_fw_version = UTIL_FW_VERSION;
360 sd_srv_version = SD_VERSION;
361 ts_srv_version = TS_VERSION;
362 hb_srv_version = HB_VERSION;
363 }
364
365 ret = vmbus_open(dev->channel, 4 * PAGE_SIZE, 4 * PAGE_SIZE, NULL, 0,
366 srv->util_cb, dev->channel);
367 if (ret)
368 goto error;
369
370 return 0;
371
372error:
373 if (srv->util_deinit)
374 srv->util_deinit();
375error1:
376 kfree(srv->recv_buffer);
377 return ret;
378}
379
380static int util_remove(struct hv_device *dev)
381{
382 struct hv_util_service *srv = hv_get_drvdata(dev);
383
384 if (srv->util_deinit)
385 srv->util_deinit();
386 vmbus_close(dev->channel);
387 kfree(srv->recv_buffer);
388
389 return 0;
390}
391
392static const struct hv_vmbus_device_id id_table[] = {
393 /* Shutdown guid */
394 { HV_SHUTDOWN_GUID,
395 .driver_data = (unsigned long)&util_shutdown
396 },
397 /* Time synch guid */
398 { HV_TS_GUID,
399 .driver_data = (unsigned long)&util_timesynch
400 },
401 /* Heartbeat guid */
402 { HV_HEART_BEAT_GUID,
403 .driver_data = (unsigned long)&util_heartbeat
404 },
405 /* KVP guid */
406 { HV_KVP_GUID,
407 .driver_data = (unsigned long)&util_kvp
408 },
409 /* VSS GUID */
410 { HV_VSS_GUID,
411 .driver_data = (unsigned long)&util_vss
412 },
413 /* File copy GUID */
414 { HV_FCOPY_GUID,
415 .driver_data = (unsigned long)&util_fcopy
416 },
417 { },
418};
419
420MODULE_DEVICE_TABLE(vmbus, id_table);
421
422/* The one and only one */
423static struct hv_driver util_drv = {
424 .name = "hv_util",
425 .id_table = id_table,
426 .probe = util_probe,
427 .remove = util_remove,
428};
429
430static int __init init_hyperv_utils(void)
431{
432 pr_info("Registering HyperV Utility Driver\n");
433
434 return vmbus_driver_register(&util_drv);
435}
436
437static void exit_hyperv_utils(void)
438{
439 pr_info("De-Registered HyperV Utility Driver\n");
440
441 vmbus_driver_unregister(&util_drv);
442}
443
444module_init(init_hyperv_utils);
445module_exit(exit_hyperv_utils);
446
447MODULE_DESCRIPTION("Hyper-V Utilities");
448MODULE_LICENSE("GPL");
1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * Copyright (c) 2010, Microsoft Corporation.
4 *
5 * Authors:
6 * Haiyang Zhang <haiyangz@microsoft.com>
7 * Hank Janssen <hjanssen@microsoft.com>
8 */
9#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
10
11#include <linux/kernel.h>
12#include <linux/init.h>
13#include <linux/module.h>
14#include <linux/slab.h>
15#include <linux/sysctl.h>
16#include <linux/reboot.h>
17#include <linux/hyperv.h>
18#include <linux/clockchips.h>
19#include <linux/ptp_clock_kernel.h>
20#include <asm/mshyperv.h>
21
22#include "hyperv_vmbus.h"
23
24#define SD_MAJOR 3
25#define SD_MINOR 0
26#define SD_MINOR_1 1
27#define SD_MINOR_2 2
28#define SD_VERSION_3_1 (SD_MAJOR << 16 | SD_MINOR_1)
29#define SD_VERSION_3_2 (SD_MAJOR << 16 | SD_MINOR_2)
30#define SD_VERSION (SD_MAJOR << 16 | SD_MINOR)
31
32#define SD_MAJOR_1 1
33#define SD_VERSION_1 (SD_MAJOR_1 << 16 | SD_MINOR)
34
35#define TS_MAJOR 4
36#define TS_MINOR 0
37#define TS_VERSION (TS_MAJOR << 16 | TS_MINOR)
38
39#define TS_MAJOR_1 1
40#define TS_VERSION_1 (TS_MAJOR_1 << 16 | TS_MINOR)
41
42#define TS_MAJOR_3 3
43#define TS_VERSION_3 (TS_MAJOR_3 << 16 | TS_MINOR)
44
45#define HB_MAJOR 3
46#define HB_MINOR 0
47#define HB_VERSION (HB_MAJOR << 16 | HB_MINOR)
48
49#define HB_MAJOR_1 1
50#define HB_VERSION_1 (HB_MAJOR_1 << 16 | HB_MINOR)
51
52static int sd_srv_version;
53static int ts_srv_version;
54static int hb_srv_version;
55
56#define SD_VER_COUNT 4
57static const int sd_versions[] = {
58 SD_VERSION_3_2,
59 SD_VERSION_3_1,
60 SD_VERSION,
61 SD_VERSION_1
62};
63
64#define TS_VER_COUNT 3
65static const int ts_versions[] = {
66 TS_VERSION,
67 TS_VERSION_3,
68 TS_VERSION_1
69};
70
71#define HB_VER_COUNT 2
72static const int hb_versions[] = {
73 HB_VERSION,
74 HB_VERSION_1
75};
76
77#define FW_VER_COUNT 2
78static const int fw_versions[] = {
79 UTIL_FW_VERSION,
80 UTIL_WS2K8_FW_VERSION
81};
82
83/*
84 * Send the "hibernate" udev event in a thread context.
85 */
86struct hibernate_work_context {
87 struct work_struct work;
88 struct hv_device *dev;
89};
90
91static struct hibernate_work_context hibernate_context;
92static bool hibernation_supported;
93
94static void send_hibernate_uevent(struct work_struct *work)
95{
96 char *uevent_env[2] = { "EVENT=hibernate", NULL };
97 struct hibernate_work_context *ctx;
98
99 ctx = container_of(work, struct hibernate_work_context, work);
100
101 kobject_uevent_env(&ctx->dev->device.kobj, KOBJ_CHANGE, uevent_env);
102
103 pr_info("Sent hibernation uevent\n");
104}
105
106static int hv_shutdown_init(struct hv_util_service *srv)
107{
108 struct vmbus_channel *channel = srv->channel;
109
110 INIT_WORK(&hibernate_context.work, send_hibernate_uevent);
111 hibernate_context.dev = channel->device_obj;
112
113 hibernation_supported = hv_is_hibernation_supported();
114
115 return 0;
116}
117
118static void shutdown_onchannelcallback(void *context);
119static struct hv_util_service util_shutdown = {
120 .util_cb = shutdown_onchannelcallback,
121 .util_init = hv_shutdown_init,
122};
123
124static int hv_timesync_init(struct hv_util_service *srv);
125static int hv_timesync_pre_suspend(void);
126static void hv_timesync_deinit(void);
127
128static void timesync_onchannelcallback(void *context);
129static struct hv_util_service util_timesynch = {
130 .util_cb = timesync_onchannelcallback,
131 .util_init = hv_timesync_init,
132 .util_pre_suspend = hv_timesync_pre_suspend,
133 .util_deinit = hv_timesync_deinit,
134};
135
136static void heartbeat_onchannelcallback(void *context);
137static struct hv_util_service util_heartbeat = {
138 .util_cb = heartbeat_onchannelcallback,
139};
140
141static struct hv_util_service util_kvp = {
142 .util_cb = hv_kvp_onchannelcallback,
143 .util_init = hv_kvp_init,
144 .util_pre_suspend = hv_kvp_pre_suspend,
145 .util_pre_resume = hv_kvp_pre_resume,
146 .util_deinit = hv_kvp_deinit,
147};
148
149static struct hv_util_service util_vss = {
150 .util_cb = hv_vss_onchannelcallback,
151 .util_init = hv_vss_init,
152 .util_pre_suspend = hv_vss_pre_suspend,
153 .util_pre_resume = hv_vss_pre_resume,
154 .util_deinit = hv_vss_deinit,
155};
156
157static struct hv_util_service util_fcopy = {
158 .util_cb = hv_fcopy_onchannelcallback,
159 .util_init = hv_fcopy_init,
160 .util_pre_suspend = hv_fcopy_pre_suspend,
161 .util_pre_resume = hv_fcopy_pre_resume,
162 .util_deinit = hv_fcopy_deinit,
163};
164
165static void perform_shutdown(struct work_struct *dummy)
166{
167 orderly_poweroff(true);
168}
169
170static void perform_restart(struct work_struct *dummy)
171{
172 orderly_reboot();
173}
174
175/*
176 * Perform the shutdown operation in a thread context.
177 */
178static DECLARE_WORK(shutdown_work, perform_shutdown);
179
180/*
181 * Perform the restart operation in a thread context.
182 */
183static DECLARE_WORK(restart_work, perform_restart);
184
185static void shutdown_onchannelcallback(void *context)
186{
187 struct vmbus_channel *channel = context;
188 struct work_struct *work = NULL;
189 u32 recvlen;
190 u64 requestid;
191 u8 *shut_txf_buf = util_shutdown.recv_buffer;
192
193 struct shutdown_msg_data *shutdown_msg;
194
195 struct icmsg_hdr *icmsghdrp;
196
197 if (vmbus_recvpacket(channel, shut_txf_buf, HV_HYP_PAGE_SIZE, &recvlen, &requestid)) {
198 pr_err_ratelimited("Shutdown request received. Could not read into shut txf buf\n");
199 return;
200 }
201
202 if (!recvlen)
203 return;
204
205 /* Ensure recvlen is big enough to read header data */
206 if (recvlen < ICMSG_HDR) {
207 pr_err_ratelimited("Shutdown request received. Packet length too small: %d\n",
208 recvlen);
209 return;
210 }
211
212 icmsghdrp = (struct icmsg_hdr *)&shut_txf_buf[sizeof(struct vmbuspipe_hdr)];
213
214 if (icmsghdrp->icmsgtype == ICMSGTYPE_NEGOTIATE) {
215 if (vmbus_prep_negotiate_resp(icmsghdrp,
216 shut_txf_buf, recvlen,
217 fw_versions, FW_VER_COUNT,
218 sd_versions, SD_VER_COUNT,
219 NULL, &sd_srv_version)) {
220 pr_info("Shutdown IC version %d.%d\n",
221 sd_srv_version >> 16,
222 sd_srv_version & 0xFFFF);
223 }
224 } else if (icmsghdrp->icmsgtype == ICMSGTYPE_SHUTDOWN) {
225 /* Ensure recvlen is big enough to contain shutdown_msg_data struct */
226 if (recvlen < ICMSG_HDR + sizeof(struct shutdown_msg_data)) {
227 pr_err_ratelimited("Invalid shutdown msg data. Packet length too small: %u\n",
228 recvlen);
229 return;
230 }
231
232 shutdown_msg = (struct shutdown_msg_data *)&shut_txf_buf[ICMSG_HDR];
233
234 /*
235 * shutdown_msg->flags can be 0(shut down), 2(reboot),
236 * or 4(hibernate). It may bitwise-OR 1, which means
237 * performing the request by force. Linux always tries
238 * to perform the request by force.
239 */
240 switch (shutdown_msg->flags) {
241 case 0:
242 case 1:
243 icmsghdrp->status = HV_S_OK;
244 work = &shutdown_work;
245 pr_info("Shutdown request received - graceful shutdown initiated\n");
246 break;
247 case 2:
248 case 3:
249 icmsghdrp->status = HV_S_OK;
250 work = &restart_work;
251 pr_info("Restart request received - graceful restart initiated\n");
252 break;
253 case 4:
254 case 5:
255 pr_info("Hibernation request received\n");
256 icmsghdrp->status = hibernation_supported ?
257 HV_S_OK : HV_E_FAIL;
258 if (hibernation_supported)
259 work = &hibernate_context.work;
260 break;
261 default:
262 icmsghdrp->status = HV_E_FAIL;
263 pr_info("Shutdown request received - Invalid request\n");
264 break;
265 }
266 } else {
267 icmsghdrp->status = HV_E_FAIL;
268 pr_err_ratelimited("Shutdown request received. Invalid msg type: %d\n",
269 icmsghdrp->icmsgtype);
270 }
271
272 icmsghdrp->icflags = ICMSGHDRFLAG_TRANSACTION
273 | ICMSGHDRFLAG_RESPONSE;
274
275 vmbus_sendpacket(channel, shut_txf_buf,
276 recvlen, requestid,
277 VM_PKT_DATA_INBAND, 0);
278
279 if (work)
280 schedule_work(work);
281}
282
283/*
284 * Set the host time in a process context.
285 */
286static struct work_struct adj_time_work;
287
288/*
289 * The last time sample, received from the host. PTP device responds to
290 * requests by using this data and the current partition-wide time reference
291 * count.
292 */
293static struct {
294 u64 host_time;
295 u64 ref_time;
296 spinlock_t lock;
297} host_ts;
298
299static inline u64 reftime_to_ns(u64 reftime)
300{
301 return (reftime - WLTIMEDELTA) * 100;
302}
303
304/*
305 * Hard coded threshold for host timesync delay: 600 seconds
306 */
307static const u64 HOST_TIMESYNC_DELAY_THRESH = 600 * (u64)NSEC_PER_SEC;
308
309static int hv_get_adj_host_time(struct timespec64 *ts)
310{
311 u64 newtime, reftime, timediff_adj;
312 unsigned long flags;
313 int ret = 0;
314
315 spin_lock_irqsave(&host_ts.lock, flags);
316 reftime = hv_read_reference_counter();
317
318 /*
319 * We need to let the caller know that last update from host
320 * is older than the max allowable threshold. clock_gettime()
321 * and PTP ioctl do not have a documented error that we could
322 * return for this specific case. Use ESTALE to report this.
323 */
324 timediff_adj = reftime - host_ts.ref_time;
325 if (timediff_adj * 100 > HOST_TIMESYNC_DELAY_THRESH) {
326 pr_warn_once("TIMESYNC IC: Stale time stamp, %llu nsecs old\n",
327 (timediff_adj * 100));
328 ret = -ESTALE;
329 }
330
331 newtime = host_ts.host_time + timediff_adj;
332 *ts = ns_to_timespec64(reftime_to_ns(newtime));
333 spin_unlock_irqrestore(&host_ts.lock, flags);
334
335 return ret;
336}
337
338static void hv_set_host_time(struct work_struct *work)
339{
340
341 struct timespec64 ts;
342
343 if (!hv_get_adj_host_time(&ts))
344 do_settimeofday64(&ts);
345}
346
347/*
348 * Synchronize time with host after reboot, restore, etc.
349 *
350 * ICTIMESYNCFLAG_SYNC flag bit indicates reboot, restore events of the VM.
351 * After reboot the flag ICTIMESYNCFLAG_SYNC is included in the first time
352 * message after the timesync channel is opened. Since the hv_utils module is
353 * loaded after hv_vmbus, the first message is usually missed. This bit is
354 * considered a hard request to discipline the clock.
355 *
356 * ICTIMESYNCFLAG_SAMPLE bit indicates a time sample from host. This is
357 * typically used as a hint to the guest. The guest is under no obligation
358 * to discipline the clock.
359 */
360static inline void adj_guesttime(u64 hosttime, u64 reftime, u8 adj_flags)
361{
362 unsigned long flags;
363 u64 cur_reftime;
364
365 /*
366 * Save the adjusted time sample from the host and the snapshot
367 * of the current system time.
368 */
369 spin_lock_irqsave(&host_ts.lock, flags);
370
371 cur_reftime = hv_read_reference_counter();
372 host_ts.host_time = hosttime;
373 host_ts.ref_time = cur_reftime;
374
375 /*
376 * TimeSync v4 messages contain reference time (guest's Hyper-V
377 * clocksource read when the time sample was generated), we can
378 * improve the precision by adding the delta between now and the
379 * time of generation. For older protocols we set
380 * reftime == cur_reftime on call.
381 */
382 host_ts.host_time += (cur_reftime - reftime);
383
384 spin_unlock_irqrestore(&host_ts.lock, flags);
385
386 /* Schedule work to do do_settimeofday64() */
387 if (adj_flags & ICTIMESYNCFLAG_SYNC)
388 schedule_work(&adj_time_work);
389}
390
391/*
392 * Time Sync Channel message handler.
393 */
394static void timesync_onchannelcallback(void *context)
395{
396 struct vmbus_channel *channel = context;
397 u32 recvlen;
398 u64 requestid;
399 struct icmsg_hdr *icmsghdrp;
400 struct ictimesync_data *timedatap;
401 struct ictimesync_ref_data *refdata;
402 u8 *time_txf_buf = util_timesynch.recv_buffer;
403
404 /*
405 * Drain the ring buffer and use the last packet to update
406 * host_ts
407 */
408 while (1) {
409 int ret = vmbus_recvpacket(channel, time_txf_buf,
410 HV_HYP_PAGE_SIZE, &recvlen,
411 &requestid);
412 if (ret) {
413 pr_err_ratelimited("TimeSync IC pkt recv failed (Err: %d)\n",
414 ret);
415 break;
416 }
417
418 if (!recvlen)
419 break;
420
421 /* Ensure recvlen is big enough to read header data */
422 if (recvlen < ICMSG_HDR) {
423 pr_err_ratelimited("Timesync request received. Packet length too small: %d\n",
424 recvlen);
425 break;
426 }
427
428 icmsghdrp = (struct icmsg_hdr *)&time_txf_buf[
429 sizeof(struct vmbuspipe_hdr)];
430
431 if (icmsghdrp->icmsgtype == ICMSGTYPE_NEGOTIATE) {
432 if (vmbus_prep_negotiate_resp(icmsghdrp,
433 time_txf_buf, recvlen,
434 fw_versions, FW_VER_COUNT,
435 ts_versions, TS_VER_COUNT,
436 NULL, &ts_srv_version)) {
437 pr_info("TimeSync IC version %d.%d\n",
438 ts_srv_version >> 16,
439 ts_srv_version & 0xFFFF);
440 }
441 } else if (icmsghdrp->icmsgtype == ICMSGTYPE_TIMESYNC) {
442 if (ts_srv_version > TS_VERSION_3) {
443 /* Ensure recvlen is big enough to read ictimesync_ref_data */
444 if (recvlen < ICMSG_HDR + sizeof(struct ictimesync_ref_data)) {
445 pr_err_ratelimited("Invalid ictimesync ref data. Length too small: %u\n",
446 recvlen);
447 break;
448 }
449 refdata = (struct ictimesync_ref_data *)&time_txf_buf[ICMSG_HDR];
450
451 adj_guesttime(refdata->parenttime,
452 refdata->vmreferencetime,
453 refdata->flags);
454 } else {
455 /* Ensure recvlen is big enough to read ictimesync_data */
456 if (recvlen < ICMSG_HDR + sizeof(struct ictimesync_data)) {
457 pr_err_ratelimited("Invalid ictimesync data. Length too small: %u\n",
458 recvlen);
459 break;
460 }
461 timedatap = (struct ictimesync_data *)&time_txf_buf[ICMSG_HDR];
462
463 adj_guesttime(timedatap->parenttime,
464 hv_read_reference_counter(),
465 timedatap->flags);
466 }
467 } else {
468 icmsghdrp->status = HV_E_FAIL;
469 pr_err_ratelimited("Timesync request received. Invalid msg type: %d\n",
470 icmsghdrp->icmsgtype);
471 }
472
473 icmsghdrp->icflags = ICMSGHDRFLAG_TRANSACTION
474 | ICMSGHDRFLAG_RESPONSE;
475
476 vmbus_sendpacket(channel, time_txf_buf,
477 recvlen, requestid,
478 VM_PKT_DATA_INBAND, 0);
479 }
480}
481
482/*
483 * Heartbeat functionality.
484 * Every two seconds, Hyper-V send us a heartbeat request message.
485 * we respond to this message, and Hyper-V knows we are alive.
486 */
487static void heartbeat_onchannelcallback(void *context)
488{
489 struct vmbus_channel *channel = context;
490 u32 recvlen;
491 u64 requestid;
492 struct icmsg_hdr *icmsghdrp;
493 struct heartbeat_msg_data *heartbeat_msg;
494 u8 *hbeat_txf_buf = util_heartbeat.recv_buffer;
495
496 while (1) {
497
498 if (vmbus_recvpacket(channel, hbeat_txf_buf, HV_HYP_PAGE_SIZE,
499 &recvlen, &requestid)) {
500 pr_err_ratelimited("Heartbeat request received. Could not read into hbeat txf buf\n");
501 return;
502 }
503
504 if (!recvlen)
505 break;
506
507 /* Ensure recvlen is big enough to read header data */
508 if (recvlen < ICMSG_HDR) {
509 pr_err_ratelimited("Heartbeat request received. Packet length too small: %d\n",
510 recvlen);
511 break;
512 }
513
514 icmsghdrp = (struct icmsg_hdr *)&hbeat_txf_buf[
515 sizeof(struct vmbuspipe_hdr)];
516
517 if (icmsghdrp->icmsgtype == ICMSGTYPE_NEGOTIATE) {
518 if (vmbus_prep_negotiate_resp(icmsghdrp,
519 hbeat_txf_buf, recvlen,
520 fw_versions, FW_VER_COUNT,
521 hb_versions, HB_VER_COUNT,
522 NULL, &hb_srv_version)) {
523
524 pr_info("Heartbeat IC version %d.%d\n",
525 hb_srv_version >> 16,
526 hb_srv_version & 0xFFFF);
527 }
528 } else if (icmsghdrp->icmsgtype == ICMSGTYPE_HEARTBEAT) {
529 /*
530 * Ensure recvlen is big enough to read seq_num. Reserved area is not
531 * included in the check as the host may not fill it up entirely
532 */
533 if (recvlen < ICMSG_HDR + sizeof(u64)) {
534 pr_err_ratelimited("Invalid heartbeat msg data. Length too small: %u\n",
535 recvlen);
536 break;
537 }
538 heartbeat_msg = (struct heartbeat_msg_data *)&hbeat_txf_buf[ICMSG_HDR];
539
540 heartbeat_msg->seq_num += 1;
541 } else {
542 icmsghdrp->status = HV_E_FAIL;
543 pr_err_ratelimited("Heartbeat request received. Invalid msg type: %d\n",
544 icmsghdrp->icmsgtype);
545 }
546
547 icmsghdrp->icflags = ICMSGHDRFLAG_TRANSACTION
548 | ICMSGHDRFLAG_RESPONSE;
549
550 vmbus_sendpacket(channel, hbeat_txf_buf,
551 recvlen, requestid,
552 VM_PKT_DATA_INBAND, 0);
553 }
554}
555
556#define HV_UTIL_RING_SEND_SIZE VMBUS_RING_SIZE(3 * HV_HYP_PAGE_SIZE)
557#define HV_UTIL_RING_RECV_SIZE VMBUS_RING_SIZE(3 * HV_HYP_PAGE_SIZE)
558
559static int util_probe(struct hv_device *dev,
560 const struct hv_vmbus_device_id *dev_id)
561{
562 struct hv_util_service *srv =
563 (struct hv_util_service *)dev_id->driver_data;
564 int ret;
565
566 srv->recv_buffer = kmalloc(HV_HYP_PAGE_SIZE * 4, GFP_KERNEL);
567 if (!srv->recv_buffer)
568 return -ENOMEM;
569 srv->channel = dev->channel;
570 if (srv->util_init) {
571 ret = srv->util_init(srv);
572 if (ret) {
573 ret = -ENODEV;
574 goto error1;
575 }
576 }
577
578 /*
579 * The set of services managed by the util driver are not performance
580 * critical and do not need batched reading. Furthermore, some services
581 * such as KVP can only handle one message from the host at a time.
582 * Turn off batched reading for all util drivers before we open the
583 * channel.
584 */
585 set_channel_read_mode(dev->channel, HV_CALL_DIRECT);
586
587 hv_set_drvdata(dev, srv);
588
589 ret = vmbus_open(dev->channel, HV_UTIL_RING_SEND_SIZE,
590 HV_UTIL_RING_RECV_SIZE, NULL, 0, srv->util_cb,
591 dev->channel);
592 if (ret)
593 goto error;
594
595 return 0;
596
597error:
598 if (srv->util_deinit)
599 srv->util_deinit();
600error1:
601 kfree(srv->recv_buffer);
602 return ret;
603}
604
605static int util_remove(struct hv_device *dev)
606{
607 struct hv_util_service *srv = hv_get_drvdata(dev);
608
609 if (srv->util_deinit)
610 srv->util_deinit();
611 vmbus_close(dev->channel);
612 kfree(srv->recv_buffer);
613
614 return 0;
615}
616
617/*
618 * When we're in util_suspend(), all the userspace processes have been frozen
619 * (refer to hibernate() -> freeze_processes()). The userspace is thawed only
620 * after the whole resume procedure, including util_resume(), finishes.
621 */
622static int util_suspend(struct hv_device *dev)
623{
624 struct hv_util_service *srv = hv_get_drvdata(dev);
625 int ret = 0;
626
627 if (srv->util_pre_suspend) {
628 ret = srv->util_pre_suspend();
629 if (ret)
630 return ret;
631 }
632
633 vmbus_close(dev->channel);
634
635 return 0;
636}
637
638static int util_resume(struct hv_device *dev)
639{
640 struct hv_util_service *srv = hv_get_drvdata(dev);
641 int ret = 0;
642
643 if (srv->util_pre_resume) {
644 ret = srv->util_pre_resume();
645 if (ret)
646 return ret;
647 }
648
649 ret = vmbus_open(dev->channel, HV_UTIL_RING_SEND_SIZE,
650 HV_UTIL_RING_RECV_SIZE, NULL, 0, srv->util_cb,
651 dev->channel);
652 return ret;
653}
654
655static const struct hv_vmbus_device_id id_table[] = {
656 /* Shutdown guid */
657 { HV_SHUTDOWN_GUID,
658 .driver_data = (unsigned long)&util_shutdown
659 },
660 /* Time synch guid */
661 { HV_TS_GUID,
662 .driver_data = (unsigned long)&util_timesynch
663 },
664 /* Heartbeat guid */
665 { HV_HEART_BEAT_GUID,
666 .driver_data = (unsigned long)&util_heartbeat
667 },
668 /* KVP guid */
669 { HV_KVP_GUID,
670 .driver_data = (unsigned long)&util_kvp
671 },
672 /* VSS GUID */
673 { HV_VSS_GUID,
674 .driver_data = (unsigned long)&util_vss
675 },
676 /* File copy GUID */
677 { HV_FCOPY_GUID,
678 .driver_data = (unsigned long)&util_fcopy
679 },
680 { },
681};
682
683MODULE_DEVICE_TABLE(vmbus, id_table);
684
685/* The one and only one */
686static struct hv_driver util_drv = {
687 .name = "hv_utils",
688 .id_table = id_table,
689 .probe = util_probe,
690 .remove = util_remove,
691 .suspend = util_suspend,
692 .resume = util_resume,
693 .driver = {
694 .probe_type = PROBE_PREFER_ASYNCHRONOUS,
695 },
696};
697
698static int hv_ptp_enable(struct ptp_clock_info *info,
699 struct ptp_clock_request *request, int on)
700{
701 return -EOPNOTSUPP;
702}
703
704static int hv_ptp_settime(struct ptp_clock_info *p, const struct timespec64 *ts)
705{
706 return -EOPNOTSUPP;
707}
708
709static int hv_ptp_adjfine(struct ptp_clock_info *ptp, long delta)
710{
711 return -EOPNOTSUPP;
712}
713static int hv_ptp_adjtime(struct ptp_clock_info *ptp, s64 delta)
714{
715 return -EOPNOTSUPP;
716}
717
718static int hv_ptp_gettime(struct ptp_clock_info *info, struct timespec64 *ts)
719{
720 return hv_get_adj_host_time(ts);
721}
722
723static struct ptp_clock_info ptp_hyperv_info = {
724 .name = "hyperv",
725 .enable = hv_ptp_enable,
726 .adjtime = hv_ptp_adjtime,
727 .adjfine = hv_ptp_adjfine,
728 .gettime64 = hv_ptp_gettime,
729 .settime64 = hv_ptp_settime,
730 .owner = THIS_MODULE,
731};
732
733static struct ptp_clock *hv_ptp_clock;
734
735static int hv_timesync_init(struct hv_util_service *srv)
736{
737 spin_lock_init(&host_ts.lock);
738
739 INIT_WORK(&adj_time_work, hv_set_host_time);
740
741 /*
742 * ptp_clock_register() returns NULL when CONFIG_PTP_1588_CLOCK is
743 * disabled but the driver is still useful without the PTP device
744 * as it still handles the ICTIMESYNCFLAG_SYNC case.
745 */
746 hv_ptp_clock = ptp_clock_register(&ptp_hyperv_info, NULL);
747 if (IS_ERR_OR_NULL(hv_ptp_clock)) {
748 pr_err("cannot register PTP clock: %d\n",
749 PTR_ERR_OR_ZERO(hv_ptp_clock));
750 hv_ptp_clock = NULL;
751 }
752
753 return 0;
754}
755
756static void hv_timesync_cancel_work(void)
757{
758 cancel_work_sync(&adj_time_work);
759}
760
761static int hv_timesync_pre_suspend(void)
762{
763 hv_timesync_cancel_work();
764 return 0;
765}
766
767static void hv_timesync_deinit(void)
768{
769 if (hv_ptp_clock)
770 ptp_clock_unregister(hv_ptp_clock);
771
772 hv_timesync_cancel_work();
773}
774
775static int __init init_hyperv_utils(void)
776{
777 pr_info("Registering HyperV Utility Driver\n");
778
779 return vmbus_driver_register(&util_drv);
780}
781
782static void exit_hyperv_utils(void)
783{
784 pr_info("De-Registered HyperV Utility Driver\n");
785
786 vmbus_driver_unregister(&util_drv);
787}
788
789module_init(init_hyperv_utils);
790module_exit(exit_hyperv_utils);
791
792MODULE_DESCRIPTION("Hyper-V Utilities");
793MODULE_LICENSE("GPL");