Release 6 errata notices for 10.3-RELEASE, all related to Microsoft Hyper-V.

Submitted by:	Dexuan Cui <decui microsoft.com>, gjb
Approved by:	so

5 files changed, 156 insertions, 49 deletions

diff --git a/sys/dev/hyperv/vmbus/hv_channel.c b/sys/dev/hyperv/vmbus/hv_channel.c
index 7037768..c144910 100644
--- a/sys/dev/hyperv/vmbus/hv_channel.c
+++ b/sys/dev/hyperv/vmbus/hv_channel.c
@@ -180,12 +180,12 @@ hv_vmbus_channel_open(
 	if (user_data_len)
 		memcpy(open_msg->user_data, user_data, user_data_len);
 
-	mtx_lock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_lock(&hv_vmbus_g_connection.channel_msg_lock);
 	TAILQ_INSERT_TAIL(
 		&hv_vmbus_g_connection.channel_msg_anchor,
 		open_info,
 		msg_list_entry);
-	mtx_unlock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_unlock(&hv_vmbus_g_connection.channel_msg_lock);
 
 	ret = hv_vmbus_post_message(
 		open_msg, sizeof(hv_vmbus_channel_open_channel));
@@ -212,12 +212,12 @@ hv_vmbus_channel_open(
 	}
 
 	cleanup:
-	mtx_lock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_lock(&hv_vmbus_g_connection.channel_msg_lock);
 	TAILQ_REMOVE(
 		&hv_vmbus_g_connection.channel_msg_anchor,
 		open_info,
 		msg_list_entry);
-	mtx_unlock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_unlock(&hv_vmbus_g_connection.channel_msg_lock);
 	sema_destroy(&open_info->wait_sema);
 	free(open_info, M_DEVBUF);
 
@@ -401,13 +401,13 @@ hv_vmbus_channel_establish_gpadl(
 	gpadl_msg->child_rel_id = channel->offer_msg.child_rel_id;
 	gpadl_msg->gpadl = next_gpadl_handle;
 
-	mtx_lock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_lock(&hv_vmbus_g_connection.channel_msg_lock);
 	TAILQ_INSERT_TAIL(
 		&hv_vmbus_g_connection.channel_msg_anchor,
 		msg_info,
 		msg_list_entry);
 
-	mtx_unlock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_unlock(&hv_vmbus_g_connection.channel_msg_lock);
 
 	ret = hv_vmbus_post_message(
 		gpadl_msg,
@@ -446,10 +446,10 @@ hv_vmbus_channel_establish_gpadl(
 
 cleanup:
 
-	mtx_lock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_lock(&hv_vmbus_g_connection.channel_msg_lock);
 	TAILQ_REMOVE(&hv_vmbus_g_connection.channel_msg_anchor,
 		msg_info, msg_list_entry);
-	mtx_unlock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_unlock(&hv_vmbus_g_connection.channel_msg_lock);
 
 	sema_destroy(&msg_info->wait_sema);
 	free(msg_info, M_DEVBUF);
@@ -488,10 +488,10 @@ hv_vmbus_channel_teardown_gpdal(
 	msg->child_rel_id = channel->offer_msg.child_rel_id;
 	msg->gpadl = gpadl_handle;
 
-	mtx_lock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_lock(&hv_vmbus_g_connection.channel_msg_lock);
 	TAILQ_INSERT_TAIL(&hv_vmbus_g_connection.channel_msg_anchor,
 			info, msg_list_entry);
-	mtx_unlock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_unlock(&hv_vmbus_g_connection.channel_msg_lock);
 
 	ret = hv_vmbus_post_message(msg,
 			sizeof(hv_vmbus_channel_gpadl_teardown));
@@ -504,10 +504,10 @@ cleanup:
 	/*
 	 * Received a torndown response
 	 */
-	mtx_lock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_lock(&hv_vmbus_g_connection.channel_msg_lock);
 	TAILQ_REMOVE(&hv_vmbus_g_connection.channel_msg_anchor,
 			info, msg_list_entry);
-	mtx_unlock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_unlock(&hv_vmbus_g_connection.channel_msg_lock);
 	sema_destroy(&info->wait_sema);
 	free(info, M_DEVBUF);
 
diff --git a/sys/dev/hyperv/vmbus/hv_channel_mgmt.c b/sys/dev/hyperv/vmbus/hv_channel_mgmt.c
index 4ccb647..a55f464 100644
--- a/sys/dev/hyperv/vmbus/hv_channel_mgmt.c
+++ b/sys/dev/hyperv/vmbus/hv_channel_mgmt.c
@@ -567,7 +567,7 @@ vmbus_channel_on_open_result(hv_vmbus_channel_msg_header* hdr)
 	/*
 	 * Find the open msg, copy the result and signal/unblock the wait event
 	 */
-	mtx_lock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_lock(&hv_vmbus_g_connection.channel_msg_lock);
 
 	TAILQ_FOREACH(msg_info, &hv_vmbus_g_connection.channel_msg_anchor,
 	    msg_list_entry) {
@@ -585,7 +585,7 @@ vmbus_channel_on_open_result(hv_vmbus_channel_msg_header* hdr)
 		}
 	    }
 	}
-	mtx_unlock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_unlock(&hv_vmbus_g_connection.channel_msg_lock);
 
 }
 
@@ -609,7 +609,7 @@ vmbus_channel_on_gpadl_created(hv_vmbus_channel_msg_header* hdr)
 	/* Find the establish msg, copy the result and signal/unblock
 	 * the wait event
 	 */
-	mtx_lock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_lock(&hv_vmbus_g_connection.channel_msg_lock);
 	TAILQ_FOREACH(msg_info, &hv_vmbus_g_connection.channel_msg_anchor,
 		msg_list_entry) {
 	    request_header = (hv_vmbus_channel_msg_header*) msg_info->msg;
@@ -628,7 +628,7 @@ vmbus_channel_on_gpadl_created(hv_vmbus_channel_msg_header* hdr)
 		}
 	    }
 	}
-	mtx_unlock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_unlock(&hv_vmbus_g_connection.channel_msg_lock);
 }
 
 /**
@@ -653,7 +653,7 @@ vmbus_channel_on_gpadl_torndown(hv_vmbus_channel_msg_header* hdr)
 	 * wait event.
 	 */
 
-	mtx_lock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_lock(&hv_vmbus_g_connection.channel_msg_lock);
 
 	TAILQ_FOREACH(msg_info, &hv_vmbus_g_connection.channel_msg_anchor,
 		msg_list_entry) {
@@ -673,7 +673,7 @@ vmbus_channel_on_gpadl_torndown(hv_vmbus_channel_msg_header* hdr)
 		}
 	    }
 	}
-    mtx_unlock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+    mtx_unlock(&hv_vmbus_g_connection.channel_msg_lock);
 }
 
 /**
@@ -693,7 +693,7 @@ vmbus_channel_on_version_response(hv_vmbus_channel_msg_header* hdr)
 
 	versionResponse = (hv_vmbus_channel_version_response*)hdr;
 
-	mtx_lock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_lock(&hv_vmbus_g_connection.channel_msg_lock);
 	TAILQ_FOREACH(msg_info, &hv_vmbus_g_connection.channel_msg_anchor,
 	    msg_list_entry) {
 	    requestHeader = (hv_vmbus_channel_msg_header*) msg_info->msg;
@@ -707,7 +707,7 @@ vmbus_channel_on_version_response(hv_vmbus_channel_msg_header* hdr)
 		sema_post(&msg_info->wait_sema);
 	    }
 	}
-    mtx_unlock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+    mtx_unlock(&hv_vmbus_g_connection.channel_msg_lock);
 
 }
 
diff --git a/sys/dev/hyperv/vmbus/hv_connection.c b/sys/dev/hyperv/vmbus/hv_connection.c
index cfdc9bb..f39616f 100644
--- a/sys/dev/hyperv/vmbus/hv_connection.c
+++ b/sys/dev/hyperv/vmbus/hv_connection.c
@@ -101,26 +101,26 @@ hv_vmbus_negotiate_version(hv_vmbus_channel_msg_info *msg_info,
 	 * Add to list before we send the request since we may receive the
 	 * response before returning from this routine
 	 */
-	mtx_lock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_lock(&hv_vmbus_g_connection.channel_msg_lock);
 
 	TAILQ_INSERT_TAIL(
 		&hv_vmbus_g_connection.channel_msg_anchor,
 		msg_info,
 		msg_list_entry);
 
-	mtx_unlock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_unlock(&hv_vmbus_g_connection.channel_msg_lock);
 
 	ret = hv_vmbus_post_message(
 		msg,
 		sizeof(hv_vmbus_channel_initiate_contact));
 
 	if (ret != 0) {
-		mtx_lock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+		mtx_lock(&hv_vmbus_g_connection.channel_msg_lock);
 		TAILQ_REMOVE(
 			&hv_vmbus_g_connection.channel_msg_anchor,
 			msg_info,
 			msg_list_entry);
-		mtx_unlock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+		mtx_unlock(&hv_vmbus_g_connection.channel_msg_lock);
 		return (ret);
 	}
 
@@ -129,12 +129,12 @@ hv_vmbus_negotiate_version(hv_vmbus_channel_msg_info *msg_info,
 	 */
 	ret = sema_timedwait(&msg_info->wait_sema, 5 * hz); /* KYS 5 seconds */
 
-	mtx_lock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_lock(&hv_vmbus_g_connection.channel_msg_lock);
 	TAILQ_REMOVE(
 		&hv_vmbus_g_connection.channel_msg_anchor,
 		msg_info,
 		msg_list_entry);
-	mtx_unlock_spin(&hv_vmbus_g_connection.channel_msg_lock);
+	mtx_unlock(&hv_vmbus_g_connection.channel_msg_lock);
 
 	/**
 	 * Check if successful
@@ -173,7 +173,7 @@ hv_vmbus_connect(void) {
 
 	TAILQ_INIT(&hv_vmbus_g_connection.channel_msg_anchor);
 	mtx_init(&hv_vmbus_g_connection.channel_msg_lock, "vmbus channel msg",
-		NULL, MTX_SPIN);
+		NULL, MTX_DEF);
 
 	TAILQ_INIT(&hv_vmbus_g_connection.channel_anchor);
 	mtx_init(&hv_vmbus_g_connection.channel_lock, "vmbus channel",
@@ -476,31 +476,35 @@ hv_vmbus_on_events(void *arg)
 /**
  * Send a msg on the vmbus's message connection
  */
-int hv_vmbus_post_message(void *buffer, size_t bufferLen) {
-	int ret = 0;
+int hv_vmbus_post_message(void *buffer, size_t bufferLen)
+{
 	hv_vmbus_connection_id connId;
-	unsigned retries = 0;
+	sbintime_t time = SBT_1MS;
+	int retries;
+	int ret;
 
-	/* NetScaler delays from previous code were consolidated here */
-	static int delayAmount[] = {100, 100, 100, 500, 500, 5000, 5000, 5000};
+	connId.as_uint32_t = 0;
+	connId.u.id = HV_VMBUS_MESSAGE_CONNECTION_ID;
 
-	/* for(each entry in delayAmount) try to post message,
-	 *  delay a little bit before retrying
+	/*
+	 * We retry to cope with transient failures caused by host side's
+	 * insufficient resources. 20 times should suffice in practice.
 	 */
-	for (retries = 0;
-	    retries < sizeof(delayAmount)/sizeof(delayAmount[0]); retries++) {
-	    connId.as_uint32_t = 0;
-	    connId.u.id = HV_VMBUS_MESSAGE_CONNECTION_ID;
-	    ret = hv_vmbus_post_msg_via_msg_ipc(connId, 1, buffer, bufferLen);
-	    if (ret != HV_STATUS_INSUFFICIENT_BUFFERS)
-		break;
-	    /* TODO: KYS We should use a blocking wait call */
-	    DELAY(delayAmount[retries]);
+	for (retries = 0; retries < 20; retries++) {
+		ret = hv_vmbus_post_msg_via_msg_ipc(connId, 1, buffer,
+						    bufferLen);
+		if (ret == HV_STATUS_SUCCESS)
+			return (0);
+
+		pause_sbt("pstmsg", time, 0, C_HARDCLOCK);
+		if (time < SBT_1S * 2)
+			time *= 2;
 	}
 
-	KASSERT(ret == 0, ("Error VMBUS: Message Post Failed\n"));
+	KASSERT(ret == HV_STATUS_SUCCESS,
+		("Error VMBUS: Message Post Failed, ret=%d\n", ret));
 
-	return (ret);
+	return (EAGAIN);
 }
 
 /**
diff --git a/sys/dev/hyperv/vmbus/hv_hv.c b/sys/dev/hyperv/vmbus/hv_hv.c
index ca5641f..f4cca18 100644
--- a/sys/dev/hyperv/vmbus/hv_hv.c
+++ b/sys/dev/hyperv/vmbus/hv_hv.c
@@ -33,6 +33,7 @@
 __FBSDID("$FreeBSD$");
 
 #include <sys/param.h>
+#include <sys/kernel.h>
 #include <sys/malloc.h>
 #include <sys/pcpu.h>
 #include <sys/timetc.h>
@@ -50,6 +51,9 @@ __FBSDID("$FreeBSD$");
 
 static u_int hv_get_timecount(struct timecounter *tc);
 
+u_int	hyperv_features;
+u_int	hyperv_recommends;
+
 /**
  * Globals
  */
@@ -211,8 +215,6 @@ hv_vmbus_init(void)
 
 	hv_vmbus_g_context.hypercall_page = virt_addr;
 
-	tc_init(&hv_timecounter); /* register virtual timecount */
-
 	hv_et_init();
 	
 	return (0);
@@ -427,3 +429,93 @@ void hv_vmbus_synic_cleanup(void *arg)
 	wrmsr(HV_X64_MSR_SIEFP, siefp.as_uint64_t);
 }
 
+static bool
+hyperv_identify(void)
+{
+	u_int regs[4];
+	unsigned int maxLeaf;
+	unsigned int op;
+
+	if (vm_guest != VM_GUEST_HV)
+		return (false);
+
+	op = HV_CPU_ID_FUNCTION_HV_VENDOR_AND_MAX_FUNCTION;
+	do_cpuid(op, regs);
+	maxLeaf = regs[0];
+	if (maxLeaf < HV_CPU_ID_FUNCTION_MS_HV_IMPLEMENTATION_LIMITS)
+		return (false);
+
+	op = HV_CPU_ID_FUNCTION_HV_INTERFACE;
+	do_cpuid(op, regs);
+	if (regs[0] != 0x31237648 /* HV#1 */)
+		return (false);
+
+	op = HV_CPU_ID_FUNCTION_MS_HV_FEATURES;
+	do_cpuid(op, regs);
+	if ((regs[0] & HV_FEATURE_MSR_HYPERCALL) == 0) {
+		/*
+		 * Hyper-V w/o Hypercall is impossible; someone
+		 * is faking Hyper-V.
+		 */
+		return (false);
+	}
+	hyperv_features = regs[0];
+
+	op = HV_CPU_ID_FUNCTION_MS_HV_VERSION;
+	do_cpuid(op, regs);
+	printf("Hyper-V Version: %d.%d.%d [SP%d]\n",
+	    regs[1] >> 16, regs[1] & 0xffff, regs[0], regs[2]);
+
+	printf("  Features: 0x%b\n", hyperv_features,
+	    "\020"
+	    "\001VPRUNTIME"
+	    "\002TMREFCNT"
+	    "\003SYNCIC"
+	    "\004SYNCTM"
+	    "\005APIC"
+	    "\006HYERCALL"
+	    "\007VPINDEX"
+	    "\010RESET"
+	    "\011STATS"
+	    "\012REFTSC"
+	    "\013IDLE"
+	    "\014TMFREQ"
+	    "\015DEBUG");
+
+	op = HV_CPU_ID_FUNCTION_MS_HV_ENLIGHTENMENT_INFORMATION;
+	do_cpuid(op, regs);
+	hyperv_recommends = regs[0];
+	if (bootverbose)
+		printf("  Recommends: %08x %08x\n", regs[0], regs[1]);
+
+	op = HV_CPU_ID_FUNCTION_MS_HV_IMPLEMENTATION_LIMITS;
+	do_cpuid(op, regs);
+	if (bootverbose) {
+		printf("  Limits: Vcpu:%d Lcpu:%d Int:%d\n",
+		    regs[0], regs[1], regs[2]);
+	}
+
+	if (maxLeaf >= HV_CPU_ID_FUNCTION_MS_HV_HARDWARE_FEATURE) {
+		op = HV_CPU_ID_FUNCTION_MS_HV_HARDWARE_FEATURE;
+		do_cpuid(op, regs);
+		if (bootverbose) {
+			printf("  HW Features: %08x AMD: %08x\n",
+			    regs[0], regs[3]);
+		}
+	}
+
+	return (true);
+}
+
+static void
+hyperv_init(void *dummy __unused)
+{
+	if (!hyperv_identify())
+		return;
+
+	if (hyperv_features & HV_FEATURE_MSR_TIME_REFCNT) {
+		/* Register virtual timecount */
+		tc_init(&hv_timecounter);
+	}
+}
+SYSINIT(hyperv_initialize, SI_SUB_HYPERVISOR, SI_ORDER_FIRST, hyperv_init, NULL);
diff --git a/sys/dev/hyperv/vmbus/hv_vmbus_priv.h b/sys/dev/hyperv/vmbus/hv_vmbus_priv.h
index 13a35c4..bf103ab 100644
--- a/sys/dev/hyperv/vmbus/hv_vmbus_priv.h
+++ b/sys/dev/hyperv/vmbus/hv_vmbus_priv.h
@@ -70,6 +70,7 @@ typedef uint16_t hv_vmbus_status;
  *    You did not supply enough message buffers to send a message.
  */
 
+#define HV_STATUS_SUCCESS                ((uint16_t)0)
 #define HV_STATUS_INSUFFICIENT_BUFFERS   ((uint16_t)0x0013)
 
 typedef void (*hv_vmbus_channel_callback)(void *context);
@@ -471,10 +472,17 @@ typedef enum {
 	HV_CPU_ID_FUNCTION_MS_HV_VERSION			= 0x40000002,
 	HV_CPU_ID_FUNCTION_MS_HV_FEATURES			= 0x40000003,
 	HV_CPU_ID_FUNCTION_MS_HV_ENLIGHTENMENT_INFORMATION	= 0x40000004,
-	HV_CPU_ID_FUNCTION_MS_HV_IMPLEMENTATION_LIMITS		= 0x40000005
-
+	HV_CPU_ID_FUNCTION_MS_HV_IMPLEMENTATION_LIMITS		= 0x40000005,
+	HV_CPU_ID_FUNCTION_MS_HV_HARDWARE_FEATURE		= 0x40000006
 } hv_vmbus_cpuid_function;
 
+#define	HV_FEATURE_MSR_TIME_REFCNT	(1 << 1)
+#define	HV_FEATURE_MSR_SYNCIC		(1 << 2)
+#define	HV_FEATURE_MSR_STIMER		(1 << 3)
+#define	HV_FEATURE_MSR_APIC		(1 << 4)
+#define	HV_FEATURE_MSR_HYPERCALL	(1 << 5)
+#define	HV_FEATURE_MSR_GUEST_IDLE	(1 << 10)
+
 /*
  * Define the format of the SIMP register
  */
@@ -628,6 +636,9 @@ typedef enum {
 extern hv_vmbus_context		hv_vmbus_g_context;
 extern hv_vmbus_connection	hv_vmbus_g_connection;
 
+extern u_int			hyperv_features;
+extern u_int			hyperv_recommends;
+
 typedef void (*vmbus_msg_handler)(hv_vmbus_channel_msg_header *msg);
 
 typedef struct hv_vmbus_channel_msg_table_entry {