2 * An implementation of key value pair (KVP) functionality for Linux.
5 * Copyright (C) 2010, Novell, Inc.
6 * Author : K. Y. Srinivasan <ksrinivasan@novell.com>
8 * This program is free software; you can redistribute it and/or modify it
9 * under the terms of the GNU General Public License version 2 as published
10 * by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
15 * NON INFRINGEMENT. See the GNU General Public License for more
18 * You should have received a copy of the GNU General Public License
19 * along with this program; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
23 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
25 #include <linux/net.h>
26 #include <linux/nls.h>
27 #include <linux/connector.h>
28 #include <linux/workqueue.h>
29 #include <linux/hyperv.h>
30 #include <asm/hyperv-tlfs.h>
32 #include "hyperv_vmbus.h"
33 #include "hv_utils_transport.h"
36 * Pre win8 version numbers used in ws2008 and ws 2008 r2 (win7)
38 #define WS2008_SRV_MAJOR 1
39 #define WS2008_SRV_MINOR 0
40 #define WS2008_SRV_VERSION (WS2008_SRV_MAJOR << 16 | WS2008_SRV_MINOR)
42 #define WIN7_SRV_MAJOR 3
43 #define WIN7_SRV_MINOR 0
44 #define WIN7_SRV_VERSION (WIN7_SRV_MAJOR << 16 | WIN7_SRV_MINOR)
46 #define WIN8_SRV_MAJOR 4
47 #define WIN8_SRV_MINOR 0
48 #define WIN8_SRV_VERSION (WIN8_SRV_MAJOR << 16 | WIN8_SRV_MINOR)
50 #define KVP_VER_COUNT 3
51 static const int kvp_versions[] = {
57 #define FW_VER_COUNT 2
58 static const int fw_versions[] = {
64 * Global state maintained for transaction that is being processed. For a class
65 * of integration services, including the "KVP service", the specified protocol
66 * is a "request/response" protocol which means that there can only be single
67 * outstanding transaction from the host at any given point in time. We use
68 * this to simplify memory management in this driver - we cache and process
69 * only one message at a time.
71 * While the request/response protocol is guaranteed by the host, we further
72 * ensure this by serializing packet processing in this driver - we do not
73 * read additional packets from the VMBUS until the current packet is fully
78 int state; /* hvutil_device_state */
79 int recv_len; /* number of bytes received. */
80 struct hv_kvp_msg *kvp_msg; /* current message */
81 struct vmbus_channel *recv_channel; /* chn we got the request */
82 u64 recv_req_id; /* request ID. */
86 * This state maintains the version number registered by the daemon.
88 static int dm_reg_value;
90 static void kvp_send_key(struct work_struct *dummy);
93 static void kvp_respond_to_host(struct hv_kvp_msg *msg, int error);
94 static void kvp_timeout_func(struct work_struct *dummy);
95 static void kvp_host_handshake_func(struct work_struct *dummy);
96 static void kvp_register(int);
98 static DECLARE_DELAYED_WORK(kvp_timeout_work, kvp_timeout_func);
99 static DECLARE_DELAYED_WORK(kvp_host_handshake_work, kvp_host_handshake_func);
100 static DECLARE_WORK(kvp_sendkey_work, kvp_send_key);
102 static const char kvp_devname[] = "vmbus/hv_kvp";
103 static u8 *recv_buffer;
104 static struct hvutil_transport *hvt;
106 * Register the kernel component with the user-level daemon.
107 * As part of this registration, pass the LIC version number.
108 * This number has no meaning, it satisfies the registration protocol.
110 #define HV_DRV_VERSION "3.1"
112 static void kvp_poll_wrapper(void *channel)
114 /* Transaction is finished, reset the state here to avoid races. */
115 kvp_transaction.state = HVUTIL_READY;
116 tasklet_schedule(&((struct vmbus_channel *)channel)->callback_event);
119 static void kvp_register_done(void)
122 * If we're still negotiating with the host cancel the timeout
123 * work to not poll the channel twice.
125 pr_debug("KVP: userspace daemon registered\n");
126 cancel_delayed_work_sync(&kvp_host_handshake_work);
127 hv_poll_channel(kvp_transaction.recv_channel, kvp_poll_wrapper);
131 kvp_register(int reg_value)
134 struct hv_kvp_msg *kvp_msg;
137 kvp_msg = kzalloc(sizeof(*kvp_msg), GFP_KERNEL);
140 version = kvp_msg->body.kvp_register.version;
141 kvp_msg->kvp_hdr.operation = reg_value;
142 strcpy(version, HV_DRV_VERSION);
144 hvutil_transport_send(hvt, kvp_msg, sizeof(*kvp_msg),
150 static void kvp_timeout_func(struct work_struct *dummy)
153 * If the timer fires, the user-mode component has not responded;
154 * process the pending transaction.
156 kvp_respond_to_host(NULL, HV_E_FAIL);
158 hv_poll_channel(kvp_transaction.recv_channel, kvp_poll_wrapper);
161 static void kvp_host_handshake_func(struct work_struct *dummy)
163 tasklet_schedule(&kvp_transaction.recv_channel->callback_event);
166 static int kvp_handle_handshake(struct hv_kvp_msg *msg)
168 switch (msg->kvp_hdr.operation) {
169 case KVP_OP_REGISTER:
170 dm_reg_value = KVP_OP_REGISTER;
171 pr_info("KVP: IP injection functionality not available\n");
172 pr_info("KVP: Upgrade the KVP daemon\n");
174 case KVP_OP_REGISTER1:
175 dm_reg_value = KVP_OP_REGISTER1;
178 pr_info("KVP: incompatible daemon\n");
179 pr_info("KVP: KVP version: %d, Daemon version: %d\n",
180 KVP_OP_REGISTER1, msg->kvp_hdr.operation);
185 * We have a compatible daemon; complete the handshake.
187 pr_debug("KVP: userspace daemon ver. %d connected\n",
188 msg->kvp_hdr.operation);
189 kvp_register(dm_reg_value);
196 * Callback when data is received from user mode.
199 static int kvp_on_msg(void *msg, int len)
201 struct hv_kvp_msg *message = (struct hv_kvp_msg *)msg;
202 struct hv_kvp_msg_enumerate *data;
205 if (len < sizeof(*message))
209 * If we are negotiating the version information
210 * with the daemon; handle that first.
213 if (kvp_transaction.state < HVUTIL_READY) {
214 return kvp_handle_handshake(message);
217 /* We didn't send anything to userspace so the reply is spurious */
218 if (kvp_transaction.state < HVUTIL_USERSPACE_REQ)
221 kvp_transaction.state = HVUTIL_USERSPACE_RECV;
224 * Based on the version of the daemon, we propagate errors from the
225 * daemon differently.
228 data = &message->body.kvp_enum_data;
230 switch (dm_reg_value) {
231 case KVP_OP_REGISTER:
233 * Null string is used to pass back error condition.
235 if (data->data.key[0] == 0)
239 case KVP_OP_REGISTER1:
241 * We use the message header information from
242 * the user level daemon to transmit errors.
244 error = message->error;
249 * Complete the transaction by forwarding the key value
250 * to the host. But first, cancel the timeout.
252 if (cancel_delayed_work_sync(&kvp_timeout_work)) {
253 kvp_respond_to_host(message, error);
254 hv_poll_channel(kvp_transaction.recv_channel, kvp_poll_wrapper);
261 static int process_ob_ipinfo(void *in_msg, void *out_msg, int op)
263 struct hv_kvp_msg *in = in_msg;
264 struct hv_kvp_ip_msg *out = out_msg;
268 case KVP_OP_GET_IP_INFO:
270 * Transform all parameters into utf16 encoding.
272 len = utf8s_to_utf16s((char *)in->body.kvp_ip_val.ip_addr,
273 strlen((char *)in->body.kvp_ip_val.ip_addr),
275 (wchar_t *)out->kvp_ip_val.ip_addr,
280 len = utf8s_to_utf16s((char *)in->body.kvp_ip_val.sub_net,
281 strlen((char *)in->body.kvp_ip_val.sub_net),
283 (wchar_t *)out->kvp_ip_val.sub_net,
288 len = utf8s_to_utf16s((char *)in->body.kvp_ip_val.gate_way,
289 strlen((char *)in->body.kvp_ip_val.gate_way),
291 (wchar_t *)out->kvp_ip_val.gate_way,
296 len = utf8s_to_utf16s((char *)in->body.kvp_ip_val.dns_addr,
297 strlen((char *)in->body.kvp_ip_val.dns_addr),
299 (wchar_t *)out->kvp_ip_val.dns_addr,
304 len = utf8s_to_utf16s((char *)in->body.kvp_ip_val.adapter_id,
305 strlen((char *)in->body.kvp_ip_val.adapter_id),
307 (wchar_t *)out->kvp_ip_val.adapter_id,
308 MAX_ADAPTER_ID_SIZE);
312 out->kvp_ip_val.dhcp_enabled =
313 in->body.kvp_ip_val.dhcp_enabled;
314 out->kvp_ip_val.addr_family =
315 in->body.kvp_ip_val.addr_family;
321 static void process_ib_ipinfo(void *in_msg, void *out_msg, int op)
323 struct hv_kvp_ip_msg *in = in_msg;
324 struct hv_kvp_msg *out = out_msg;
327 case KVP_OP_SET_IP_INFO:
329 * Transform all parameters into utf8 encoding.
331 utf16s_to_utf8s((wchar_t *)in->kvp_ip_val.ip_addr,
334 (__u8 *)out->body.kvp_ip_val.ip_addr,
337 utf16s_to_utf8s((wchar_t *)in->kvp_ip_val.sub_net,
340 (__u8 *)out->body.kvp_ip_val.sub_net,
343 utf16s_to_utf8s((wchar_t *)in->kvp_ip_val.gate_way,
346 (__u8 *)out->body.kvp_ip_val.gate_way,
349 utf16s_to_utf8s((wchar_t *)in->kvp_ip_val.dns_addr,
352 (__u8 *)out->body.kvp_ip_val.dns_addr,
355 out->body.kvp_ip_val.dhcp_enabled = in->kvp_ip_val.dhcp_enabled;
359 case KVP_OP_GET_IP_INFO:
360 utf16s_to_utf8s((wchar_t *)in->kvp_ip_val.adapter_id,
363 (__u8 *)out->body.kvp_ip_val.adapter_id,
364 MAX_ADAPTER_ID_SIZE);
366 out->body.kvp_ip_val.addr_family = in->kvp_ip_val.addr_family;
374 kvp_send_key(struct work_struct *dummy)
376 struct hv_kvp_msg *message;
377 struct hv_kvp_msg *in_msg;
378 __u8 operation = kvp_transaction.kvp_msg->kvp_hdr.operation;
379 __u8 pool = kvp_transaction.kvp_msg->kvp_hdr.pool;
384 /* The transaction state is wrong. */
385 if (kvp_transaction.state != HVUTIL_HOSTMSG_RECEIVED)
388 message = kzalloc(sizeof(*message), GFP_KERNEL);
392 message->kvp_hdr.operation = operation;
393 message->kvp_hdr.pool = pool;
394 in_msg = kvp_transaction.kvp_msg;
397 * The key/value strings sent from the host are encoded in
398 * in utf16; convert it to utf8 strings.
399 * The host assures us that the utf16 strings will not exceed
400 * the max lengths specified. We will however, reserve room
401 * for the string terminating character - in the utf16s_utf8s()
402 * function we limit the size of the buffer where the converted
403 * string is placed to HV_KVP_EXCHANGE_MAX_*_SIZE -1 to guarantee
404 * that the strings can be properly terminated!
407 switch (message->kvp_hdr.operation) {
408 case KVP_OP_SET_IP_INFO:
409 process_ib_ipinfo(in_msg, message, KVP_OP_SET_IP_INFO);
411 case KVP_OP_GET_IP_INFO:
413 * We only need to pass on the info of operation, adapter_id
414 * and addr_family to the userland kvp daemon.
416 process_ib_ipinfo(in_msg, message, KVP_OP_GET_IP_INFO);
419 switch (in_msg->body.kvp_set.data.value_type) {
422 * The value is a string - utf16 encoding.
424 message->body.kvp_set.data.value_size =
426 (wchar_t *)in_msg->body.kvp_set.data.value,
427 in_msg->body.kvp_set.data.value_size,
429 message->body.kvp_set.data.value,
430 HV_KVP_EXCHANGE_MAX_VALUE_SIZE - 1) + 1;
435 * The value is a 32 bit scalar.
436 * We save this as a utf8 string.
438 val32 = in_msg->body.kvp_set.data.value_u32;
439 message->body.kvp_set.data.value_size =
440 sprintf(message->body.kvp_set.data.value,
446 * The value is a 64 bit scalar.
447 * We save this as a utf8 string.
449 val64 = in_msg->body.kvp_set.data.value_u64;
450 message->body.kvp_set.data.value_size =
451 sprintf(message->body.kvp_set.data.value,
458 * The key is always a string - utf16 encoding.
460 message->body.kvp_set.data.key_size =
462 (wchar_t *)in_msg->body.kvp_set.data.key,
463 in_msg->body.kvp_set.data.key_size,
465 message->body.kvp_set.data.key,
466 HV_KVP_EXCHANGE_MAX_KEY_SIZE - 1) + 1;
471 message->body.kvp_get.data.key_size =
473 (wchar_t *)in_msg->body.kvp_get.data.key,
474 in_msg->body.kvp_get.data.key_size,
476 message->body.kvp_get.data.key,
477 HV_KVP_EXCHANGE_MAX_KEY_SIZE - 1) + 1;
481 message->body.kvp_delete.key_size =
483 (wchar_t *)in_msg->body.kvp_delete.key,
484 in_msg->body.kvp_delete.key_size,
486 message->body.kvp_delete.key,
487 HV_KVP_EXCHANGE_MAX_KEY_SIZE - 1) + 1;
490 case KVP_OP_ENUMERATE:
491 message->body.kvp_enum_data.index =
492 in_msg->body.kvp_enum_data.index;
496 kvp_transaction.state = HVUTIL_USERSPACE_REQ;
497 rc = hvutil_transport_send(hvt, message, sizeof(*message), NULL);
499 pr_debug("KVP: failed to communicate to the daemon: %d\n", rc);
500 if (cancel_delayed_work_sync(&kvp_timeout_work)) {
501 kvp_respond_to_host(message, HV_E_FAIL);
502 kvp_transaction.state = HVUTIL_READY;
510 * Send a response back to the host.
514 kvp_respond_to_host(struct hv_kvp_msg *msg_to_host, int error)
516 struct hv_kvp_msg *kvp_msg;
517 struct hv_kvp_exchg_msg_value *kvp_data;
520 struct icmsg_hdr *icmsghdrp;
524 struct vmbus_channel *channel;
529 * Copy the global state for completing the transaction. Note that
530 * only one transaction can be active at a time.
533 buf_len = kvp_transaction.recv_len;
534 channel = kvp_transaction.recv_channel;
535 req_id = kvp_transaction.recv_req_id;
537 icmsghdrp = (struct icmsg_hdr *)
538 &recv_buffer[sizeof(struct vmbuspipe_hdr)];
540 if (channel->onchannel_callback == NULL)
542 * We have raced with util driver being unloaded;
547 icmsghdrp->status = error;
550 * If the error parameter is set, terminate the host's enumeration
555 * Something failed or we have timed out;
556 * terminate the current host-side iteration.
561 kvp_msg = (struct hv_kvp_msg *)
562 &recv_buffer[sizeof(struct vmbuspipe_hdr) +
563 sizeof(struct icmsg_hdr)];
565 switch (kvp_transaction.kvp_msg->kvp_hdr.operation) {
566 case KVP_OP_GET_IP_INFO:
567 ret = process_ob_ipinfo(msg_to_host,
568 (struct hv_kvp_ip_msg *)kvp_msg,
571 icmsghdrp->status = HV_E_FAIL;
574 case KVP_OP_SET_IP_INFO:
577 kvp_data = &kvp_msg->body.kvp_get.data;
588 kvp_data = &kvp_msg->body.kvp_enum_data.data;
589 key_name = msg_to_host->body.kvp_enum_data.data.key;
592 * The windows host expects the key/value pair to be encoded
593 * in utf16. Ensure that the key/value size reported to the host
594 * will be less than or equal to the MAX size (including the
595 * terminating character).
597 keylen = utf8s_to_utf16s(key_name, strlen(key_name), UTF16_HOST_ENDIAN,
598 (wchar_t *) kvp_data->key,
599 (HV_KVP_EXCHANGE_MAX_KEY_SIZE / 2) - 2);
600 kvp_data->key_size = 2*(keylen + 1); /* utf16 encoding */
603 value = msg_to_host->body.kvp_enum_data.data.value;
604 valuelen = utf8s_to_utf16s(value, strlen(value), UTF16_HOST_ENDIAN,
605 (wchar_t *) kvp_data->value,
606 (HV_KVP_EXCHANGE_MAX_VALUE_SIZE / 2) - 2);
607 kvp_data->value_size = 2*(valuelen + 1); /* utf16 encoding */
610 * If the utf8s to utf16s conversion failed; notify host
613 if ((keylen < 0) || (valuelen < 0))
614 icmsghdrp->status = HV_E_FAIL;
616 kvp_data->value_type = REG_SZ; /* all our values are strings */
619 icmsghdrp->icflags = ICMSGHDRFLAG_TRANSACTION | ICMSGHDRFLAG_RESPONSE;
621 vmbus_sendpacket(channel, recv_buffer, buf_len, req_id,
622 VM_PKT_DATA_INBAND, 0);
626 * This callback is invoked when we get a KVP message from the host.
627 * The host ensures that only one KVP transaction can be active at a time.
628 * KVP implementation in Linux needs to forward the key to a user-mde
629 * component to retrieve the corresponding value. Consequently, we cannot
630 * respond to the host in the context of this callback. Since the host
631 * guarantees that at most only one transaction can be active at a time,
632 * we stash away the transaction state in a set of global variables.
635 void hv_kvp_onchannelcallback(void *context)
637 struct vmbus_channel *channel = context;
641 struct hv_kvp_msg *kvp_msg;
643 struct icmsg_hdr *icmsghdrp;
645 static enum {NEGO_NOT_STARTED,
647 NEGO_FINISHED} host_negotiatied = NEGO_NOT_STARTED;
649 if (kvp_transaction.state < HVUTIL_READY) {
651 * If userspace daemon is not connected and host is asking
652 * us to negotiate we need to delay to not lose messages.
653 * This is important for Failover IP setting.
655 if (host_negotiatied == NEGO_NOT_STARTED) {
656 host_negotiatied = NEGO_IN_PROGRESS;
657 schedule_delayed_work(&kvp_host_handshake_work,
658 HV_UTIL_NEGO_TIMEOUT * HZ);
662 if (kvp_transaction.state > HVUTIL_READY)
665 if (vmbus_recvpacket(channel, recv_buffer, HV_HYP_PAGE_SIZE * 4, &recvlen, &requestid)) {
666 pr_err_ratelimited("KVP request received. Could not read into recv buf\n");
673 /* Ensure recvlen is big enough to read header data */
674 if (recvlen < ICMSG_HDR) {
675 pr_err_ratelimited("KVP request received. Packet length too small: %d\n",
680 icmsghdrp = (struct icmsg_hdr *)&recv_buffer[sizeof(struct vmbuspipe_hdr)];
682 if (icmsghdrp->icmsgtype == ICMSGTYPE_NEGOTIATE) {
683 if (vmbus_prep_negotiate_resp(icmsghdrp,
684 recv_buffer, recvlen,
685 fw_versions, FW_VER_COUNT,
686 kvp_versions, KVP_VER_COUNT,
687 NULL, &kvp_srv_version)) {
688 pr_info("KVP IC version %d.%d\n",
689 kvp_srv_version >> 16,
690 kvp_srv_version & 0xFFFF);
692 } else if (icmsghdrp->icmsgtype == ICMSGTYPE_KVPEXCHANGE) {
694 * recvlen is not checked against sizeof(struct kvp_msg) because kvp_msg contains
695 * a union of structs and the msg type received is not known. Code using this
696 * struct should provide validation when accessing its fields.
698 kvp_msg = (struct hv_kvp_msg *)&recv_buffer[ICMSG_HDR];
701 * Stash away this global state for completing the
702 * transaction; note transactions are serialized.
705 kvp_transaction.recv_len = recvlen;
706 kvp_transaction.recv_req_id = requestid;
707 kvp_transaction.kvp_msg = kvp_msg;
709 if (kvp_transaction.state < HVUTIL_READY) {
710 /* Userspace is not registered yet */
711 kvp_respond_to_host(NULL, HV_E_FAIL);
714 kvp_transaction.state = HVUTIL_HOSTMSG_RECEIVED;
717 * Get the information from the
718 * user-mode component.
719 * component. This transaction will be
720 * completed when we get the value from
721 * the user-mode component.
722 * Set a timeout to deal with
723 * user-mode not responding.
725 schedule_work(&kvp_sendkey_work);
726 schedule_delayed_work(&kvp_timeout_work,
727 HV_UTIL_TIMEOUT * HZ);
732 pr_err_ratelimited("KVP request received. Invalid msg type: %d\n",
733 icmsghdrp->icmsgtype);
737 icmsghdrp->icflags = ICMSGHDRFLAG_TRANSACTION
738 | ICMSGHDRFLAG_RESPONSE;
740 vmbus_sendpacket(channel, recv_buffer,
742 VM_PKT_DATA_INBAND, 0);
744 host_negotiatied = NEGO_FINISHED;
745 hv_poll_channel(kvp_transaction.recv_channel, kvp_poll_wrapper);
748 static void kvp_on_reset(void)
750 if (cancel_delayed_work_sync(&kvp_timeout_work))
751 kvp_respond_to_host(NULL, HV_E_FAIL);
752 kvp_transaction.state = HVUTIL_DEVICE_INIT;
756 hv_kvp_init(struct hv_util_service *srv)
758 recv_buffer = srv->recv_buffer;
759 kvp_transaction.recv_channel = srv->channel;
760 kvp_transaction.recv_channel->max_pkt_size = HV_HYP_PAGE_SIZE * 4;
763 * When this driver loads, the user level daemon that
764 * processes the host requests may not yet be running.
765 * Defer processing channel callbacks until the daemon
768 kvp_transaction.state = HVUTIL_DEVICE_INIT;
770 hvt = hvutil_transport_init(kvp_devname, CN_KVP_IDX, CN_KVP_VAL,
771 kvp_on_msg, kvp_on_reset);
778 static void hv_kvp_cancel_work(void)
780 cancel_delayed_work_sync(&kvp_host_handshake_work);
781 cancel_delayed_work_sync(&kvp_timeout_work);
782 cancel_work_sync(&kvp_sendkey_work);
785 int hv_kvp_pre_suspend(void)
787 struct vmbus_channel *channel = kvp_transaction.recv_channel;
789 tasklet_disable(&channel->callback_event);
792 * If there is a pending transtion, it's unnecessary to tell the host
793 * that the transaction will fail, because that is implied when
794 * util_suspend() calls vmbus_close() later.
796 hv_kvp_cancel_work();
799 * Forece the state to READY to handle the ICMSGTYPE_NEGOTIATE message
800 * later. The user space daemon may go out of order and its write()
801 * may fail with EINVAL: this doesn't matter since the daemon will
802 * reset the device by closing and re-opening it.
804 kvp_transaction.state = HVUTIL_READY;
808 int hv_kvp_pre_resume(void)
810 struct vmbus_channel *channel = kvp_transaction.recv_channel;
812 tasklet_enable(&channel->callback_event);
817 void hv_kvp_deinit(void)
819 kvp_transaction.state = HVUTIL_DEVICE_DYING;
821 hv_kvp_cancel_work();
823 hvutil_transport_destroy(hvt);