Merge tag 'for-linus-5.7-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rw/uml

Pull UML updates from Richard Weinberger:

 - New mode for time travel, external via virtio

 - Fixes for ubd to make sure no requests can get lost

 - Fixes for vector networking

 - Allow CONFIG_STATIC_LINK only when possible

 - Minor cleanups and fixes

* tag 'for-linus-5.7-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/rw/uml:
  um: Remove some unnecessary NULL checks in vector_user.c
  um: vector: Avoid NULL ptr deference if transport is unset
  um: Make CONFIG_STATIC_LINK actually static
  um: Implement cpu_relax() as ndelay(1) for time-travel
  um: Implement ndelay/udelay in time-travel mode
  um: Implement time-travel=ext
  um: virtio: Implement VHOST_USER_PROTOCOL_F_INBAND_NOTIFICATIONS
  um: time-travel: Rewrite as an event scheduler
  um: Move timer-internal.h to non-shared
  hostfs: Use kasprintf() instead of fixed buffer formatting
  um: falloc.h needs to be directly included for older libc
  um: ubd: Retry buffer read on any kind of error
  um: ubd: Prevent buffer overrun on command completion
  um: Fix overlapping ELF segments when statically linked
  um: Delete never executed timer
  um: Don't overwrite ethtool driver version
  um: Fix len of file in create_pid_file
  um: Don't use console_drivers directly
  um: Cleanup CONFIG_IOSCHED_CFQ

24 files changed, 854 insertions, 206 deletions

diff --git a/arch/um/Kconfig b/arch/um/Kconfig
index 0917f8443c28..96ab7026b037 100644
--- a/arch/um/Kconfig
+++ b/arch/um/Kconfig
@@ -62,9 +62,12 @@ config NR_CPUS
 
 source "arch/$(HEADER_ARCH)/um/Kconfig"
 
+config FORBID_STATIC_LINK
+	bool
+
 config STATIC_LINK
 	bool "Force a static link"
-	default n
+	depends on !FORBID_STATIC_LINK
 	help
 	  This option gives you the ability to force a static link of UML.
 	  Normally, UML is linked as a shared binary.  This is inconvenient for
@@ -73,6 +76,9 @@ config STATIC_LINK
 	  Additionally, this option enables using higher memory spaces (up to
 	  2.75G) for UML.
 
+	  NOTE: This option is incompatible with some networking features which
+	  depend on features that require being dynamically loaded (like NSS).
+
 config LD_SCRIPT_STATIC
 	bool
 	default y
@@ -191,6 +197,7 @@ config UML_TIME_TRAVEL_SUPPORT
 	prompt "Support time-travel mode (e.g. for test execution)"
 	# inf-cpu mode is incompatible with the benchmarking
 	depends on !RAID6_PQ_BENCHMARK
+	depends on !SMP
 	help
 	  Enable this option to support time travel inside the UML instance.
 
diff --git a/arch/um/configs/i386_defconfig b/arch/um/configs/i386_defconfig
index 73e98bb57bf5..fb51bd206dbe 100644
--- a/arch/um/configs/i386_defconfig
+++ b/arch/um/configs/i386_defconfig
@@ -26,7 +26,7 @@ CONFIG_SLAB=y
 CONFIG_MODULES=y
 CONFIG_MODULE_UNLOAD=y
 # CONFIG_BLK_DEV_BSG is not set
-CONFIG_IOSCHED_CFQ=m
+CONFIG_IOSCHED_BFQ=m
 CONFIG_SSL=y
 CONFIG_NULL_CHAN=y
 CONFIG_PORT_CHAN=y
diff --git a/arch/um/configs/x86_64_defconfig b/arch/um/configs/x86_64_defconfig
index 3281d7600225..477b87317424 100644
--- a/arch/um/configs/x86_64_defconfig
+++ b/arch/um/configs/x86_64_defconfig
@@ -24,7 +24,7 @@ CONFIG_SLAB=y
 CONFIG_MODULES=y
 CONFIG_MODULE_UNLOAD=y
 # CONFIG_BLK_DEV_BSG is not set
-CONFIG_IOSCHED_CFQ=m
+CONFIG_IOSCHED_BFQ=m
 CONFIG_SSL=y
 CONFIG_NULL_CHAN=y
 CONFIG_PORT_CHAN=y
diff --git a/arch/um/drivers/Kconfig b/arch/um/drivers/Kconfig
index 72d417055782..9160ead56e33 100644
--- a/arch/um/drivers/Kconfig
+++ b/arch/um/drivers/Kconfig
@@ -234,6 +234,7 @@ config UML_NET_DAEMON
 config UML_NET_VECTOR
 	bool "Vector I/O high performance network devices"
 	depends on UML_NET
+	select FORBID_STATIC_LINK
 	help
 	This User-Mode Linux network driver uses multi-message send
 	and receive functions. The host running the UML guest must have
@@ -245,6 +246,7 @@ config UML_NET_VECTOR
 config UML_NET_VDE
 	bool "VDE transport (obsolete)"
 	depends on UML_NET
+	select FORBID_STATIC_LINK
 	help
 	This User-Mode Linux network transport allows one or more running
 	UMLs on a single host to communicate with each other and also
@@ -292,6 +294,7 @@ config UML_NET_MCAST
 config UML_NET_PCAP
 	bool "pcap transport (obsolete)"
 	depends on UML_NET
+	select FORBID_STATIC_LINK
 	help
 	The pcap transport makes a pcap packet stream on the host look
 	like an ethernet device inside UML.  This is useful for making
diff --git a/arch/um/drivers/net_kern.c b/arch/um/drivers/net_kern.c
index 35ebeebfc1a8..1802cf4ef5a5 100644
--- a/arch/um/drivers/net_kern.c
+++ b/arch/um/drivers/net_kern.c
@@ -266,7 +266,6 @@ static void uml_net_get_drvinfo(struct net_device *dev,
 				struct ethtool_drvinfo *info)
 {
 	strlcpy(info->driver, DRIVER_NAME, sizeof(info->driver));
-	strlcpy(info->version, "42", sizeof(info->version));
 }
 
 static const struct ethtool_ops uml_net_ethtool_ops = {
@@ -275,17 +274,6 @@ static const struct ethtool_ops uml_net_ethtool_ops = {
 	.get_ts_info	= ethtool_op_get_ts_info,
 };
 
-static void uml_net_user_timer_expire(struct timer_list *t)
-{
-#ifdef undef
-	struct uml_net_private *lp = from_timer(lp, t, tl);
-	struct connection *conn = &lp->user;
-
-	dprintk(KERN_INFO "uml_net_user_timer_expire [%p]\n", conn);
-	do_connect(conn);
-#endif
-}
-
 void uml_net_setup_etheraddr(struct net_device *dev, char *str)
 {
 	unsigned char *addr = dev->dev_addr;
@@ -456,7 +444,6 @@ static void eth_configure(int n, void *init, char *mac,
 		  .add_address 		= transport->user->add_address,
 		  .delete_address  	= transport->user->delete_address });
 
-	timer_setup(&lp->tl, uml_net_user_timer_expire, 0);
 	spin_lock_init(&lp->lock);
 	memcpy(lp->mac, dev->dev_addr, sizeof(lp->mac));
 
diff --git a/arch/um/drivers/ubd_kern.c b/arch/um/drivers/ubd_kern.c
index 247f95da057b..eae8c83364f7 100644
--- a/arch/um/drivers/ubd_kern.c
+++ b/arch/um/drivers/ubd_kern.c
@@ -1592,11 +1592,11 @@ int io_thread(void *arg)
 			&io_remainder_size,
 			UBD_REQ_BUFFER_SIZE
 		);
-		if (n < 0) {
-			if (n == -EAGAIN) {
+		if (n <= 0) {
+			if (n == -EAGAIN)
 				ubd_read_poll(-1);
-				continue;
-			}
+
+			continue;
 		}
 
 		for (count = 0; count < n/sizeof(struct io_thread_req *); count++) {
@@ -1607,7 +1607,9 @@ int io_thread(void *arg)
 		written = 0;
 
 		do {
-			res = os_write_file(kernel_fd, ((char *) io_req_buffer) + written, n);
+			res = os_write_file(kernel_fd,
+					    ((char *) io_req_buffer) + written,
+					    n - written);
 			if (res >= 0) {
 				written += res;
 			}
diff --git a/arch/um/drivers/vector_kern.c b/arch/um/drivers/vector_kern.c
index e98304d0219e..8735c468230a 100644
--- a/arch/um/drivers/vector_kern.c
+++ b/arch/um/drivers/vector_kern.c
@@ -46,7 +46,6 @@
 
 
 #define DRIVER_NAME "uml-vector"
-#define DRIVER_VERSION "01"
 struct vector_cmd_line_arg {
 	struct list_head list;
 	int unit;
@@ -198,6 +197,9 @@ static int get_transport_options(struct arglist *def)
 	long parsed;
 	int result = 0;
 
+	if (transport == NULL)
+		return -EINVAL;
+
 	if (vector != NULL) {
 		if (kstrtoul(vector, 10, &parsed) == 0) {
 			if (parsed == 0) {
@@ -1378,7 +1380,6 @@ static void vector_net_get_drvinfo(struct net_device *dev,
 				struct ethtool_drvinfo *info)
 {
 	strlcpy(info->driver, DRIVER_NAME, sizeof(info->driver));
-	strlcpy(info->version, DRIVER_VERSION, sizeof(info->version));
 }
 
 static int vector_net_load_bpf_flash(struct net_device *dev,
diff --git a/arch/um/drivers/vector_user.c b/arch/um/drivers/vector_user.c
index ddcd917be0af..aa28e9eecb7b 100644
--- a/arch/um/drivers/vector_user.c
+++ b/arch/um/drivers/vector_user.c
@@ -221,8 +221,7 @@ static struct vector_fds *user_init_tap_fds(struct arglist *ifspec)
 	return result;
 tap_cleanup:
 	printk(UM_KERN_ERR "user_init_tap: init failed, error %d", fd);
-	if (result != NULL)
-		kfree(result);
+	kfree(result);
 	return NULL;
 }
 
@@ -266,8 +265,7 @@ static struct vector_fds *user_init_hybrid_fds(struct arglist *ifspec)
 	return result;
 hybrid_cleanup:
 	printk(UM_KERN_ERR "user_init_hybrid: init failed");
-	if (result != NULL)
-		kfree(result);
+	kfree(result);
 	return NULL;
 }
 
@@ -344,10 +342,8 @@ static struct vector_fds *user_init_unix_fds(struct arglist *ifspec, int id)
 unix_cleanup:
 	if (fd >= 0)
 		os_close_file(fd);
-	if (remote_addr != NULL)
-		kfree(remote_addr);
-	if (result != NULL)
-		kfree(result);
+	kfree(remote_addr);
+	kfree(result);
 	return NULL;
 }
 
@@ -382,8 +378,7 @@ static struct vector_fds *user_init_raw_fds(struct arglist *ifspec)
 	return result;
 raw_cleanup:
 	printk(UM_KERN_ERR "user_init_raw: init failed, error %d", err);
-	if (result != NULL)
-		kfree(result);
+	kfree(result);
 	return NULL;
 }
 
diff --git a/arch/um/drivers/vhost_user.h b/arch/um/drivers/vhost_user.h
index 45ff5ea22fea..6c71b6005177 100644
--- a/arch/um/drivers/vhost_user.h
+++ b/arch/um/drivers/vhost_user.h
@@ -10,9 +10,10 @@
 /* Feature bits */
 #define VHOST_USER_F_PROTOCOL_FEATURES	30
 /* Protocol feature bits */
-#define VHOST_USER_PROTOCOL_F_REPLY_ACK		3
-#define VHOST_USER_PROTOCOL_F_SLAVE_REQ		5
-#define VHOST_USER_PROTOCOL_F_CONFIG		9
+#define VHOST_USER_PROTOCOL_F_REPLY_ACK			3
+#define VHOST_USER_PROTOCOL_F_SLAVE_REQ			5
+#define VHOST_USER_PROTOCOL_F_CONFIG			9
+#define VHOST_USER_PROTOCOL_F_INBAND_NOTIFICATIONS	14
 /* Vring state index masks */
 #define VHOST_USER_VRING_INDEX_MASK	0xff
 #define VHOST_USER_VRING_POLL_MASK	BIT(8)
@@ -24,7 +25,8 @@
 /* Supported protocol features */
 #define VHOST_USER_SUPPORTED_PROTOCOL_F	(BIT_ULL(VHOST_USER_PROTOCOL_F_REPLY_ACK) | \
 					 BIT_ULL(VHOST_USER_PROTOCOL_F_SLAVE_REQ) | \
-					 BIT_ULL(VHOST_USER_PROTOCOL_F_CONFIG))
+					 BIT_ULL(VHOST_USER_PROTOCOL_F_CONFIG) | \
+					 BIT_ULL(VHOST_USER_PROTOCOL_F_INBAND_NOTIFICATIONS))
 
 enum vhost_user_request {
 	VHOST_USER_GET_FEATURES = 1,
@@ -52,12 +54,14 @@ enum vhost_user_request {
 	VHOST_USER_SET_VRING_ENDIAN = 23,
 	VHOST_USER_GET_CONFIG = 24,
 	VHOST_USER_SET_CONFIG = 25,
+	VHOST_USER_VRING_KICK = 35,
 };
 
 enum vhost_user_slave_request {
 	VHOST_USER_SLAVE_IOTLB_MSG = 1,
 	VHOST_USER_SLAVE_CONFIG_CHANGE_MSG = 2,
 	VHOST_USER_SLAVE_VRING_HOST_NOTIFIER_MSG = 3,
+	VHOST_USER_SLAVE_VRING_CALL = 4,
 };
 
 struct vhost_user_header {
diff --git a/arch/um/drivers/virtio_uml.c b/arch/um/drivers/virtio_uml.c
index 023ced2250ea..be54d368e73d 100644
--- a/arch/um/drivers/virtio_uml.c
+++ b/arch/um/drivers/virtio_uml.c
@@ -26,6 +26,7 @@
 #include <linux/virtio.h>
 #include <linux/virtio_config.h>
 #include <linux/virtio_ring.h>
+#include <linux/time-internal.h>
 #include <shared/as-layout.h>
 #include <irq_kern.h>
 #include <init.h>
@@ -53,6 +54,7 @@ struct virtio_uml_device {
 	struct virtio_device vdev;
 	struct platform_device *pdev;
 
+	spinlock_t sock_lock;
 	int sock, req_fd;
 	u64 features;
 	u64 protocol_features;
@@ -63,6 +65,11 @@ struct virtio_uml_device {
 struct virtio_uml_vq_info {
 	int kick_fd, call_fd;
 	char name[32];
+#ifdef CONFIG_UML_TIME_TRAVEL_SUPPORT
+	struct virtqueue *vq;
+	vq_callback_t *callback;
+	struct time_travel_event defer;
+#endif
 };
 
 extern unsigned long long physmem_size, highmem;
@@ -117,10 +124,27 @@ static int vhost_user_recv_header(int fd, struct vhost_user_msg *msg)
 
 static int vhost_user_recv(struct virtio_uml_device *vu_dev,
 			   int fd, struct vhost_user_msg *msg,
-			   size_t max_payload_size)
+			   size_t max_payload_size, bool wait)
 {
 	size_t size;
-	int rc = vhost_user_recv_header(fd, msg);
+	int rc;
+
+	/*
+	 * In virtio time-travel mode, we're handling all the vhost-user
+	 * FDs by polling them whenever appropriate. However, we may get
+	 * into a situation where we're sending out an interrupt message
+	 * to a device (e.g. a net device) and need to handle a simulation
+	 * time message while doing so, e.g. one that tells us to update
+	 * our idea of how long we can run without scheduling.
+	 *
+	 * Thus, we need to not just read() from the given fd, but need
+	 * to also handle messages for the simulation time - this function
+	 * does that for us while waiting for the given fd to be readable.
+	 */
+	if (wait)
+		time_travel_wait_readable(fd);
+
+	rc = vhost_user_recv_header(fd, msg);
 
 	if (rc == -ECONNRESET && vu_dev->registered) {
 		struct virtio_uml_platform_data *pdata;
@@ -142,7 +166,8 @@ static int vhost_user_recv_resp(struct virtio_uml_device *vu_dev,
 				struct vhost_user_msg *msg,
 				size_t max_payload_size)
 {
-	int rc = vhost_user_recv(vu_dev, vu_dev->sock, msg, max_payload_size);
+	int rc = vhost_user_recv(vu_dev, vu_dev->sock, msg,
+				 max_payload_size, true);
 
 	if (rc)
 		return rc;
@@ -172,7 +197,8 @@ static int vhost_user_recv_req(struct virtio_uml_device *vu_dev,
 			       struct vhost_user_msg *msg,
 			       size_t max_payload_size)
 {
-	int rc = vhost_user_recv(vu_dev, vu_dev->req_fd, msg, max_payload_size);
+	int rc = vhost_user_recv(vu_dev, vu_dev->req_fd, msg,
+				 max_payload_size, false);
 
 	if (rc)
 		return rc;
@@ -189,6 +215,7 @@ static int vhost_user_send(struct virtio_uml_device *vu_dev,
 			   int *fds, size_t num_fds)
 {
 	size_t size = sizeof(msg->header) + msg->header.size;
+	unsigned long flags;
 	bool request_ack;
 	int rc;
 
@@ -207,24 +234,28 @@ static int vhost_user_send(struct virtio_uml_device *vu_dev,
 	if (request_ack)
 		msg->header.flags |= VHOST_USER_FLAG_NEED_REPLY;
 
+	spin_lock_irqsave(&vu_dev->sock_lock, flags);
 	rc = full_sendmsg_fds(vu_dev->sock, msg, size, fds, num_fds);
 	if (rc < 0)
-		return rc;
+		goto out;
 
 	if (request_ack) {
 		uint64_t status;
 
 		rc = vhost_user_recv_u64(vu_dev, &status);
 		if (rc)
-			return rc;
+			goto out;
 
 		if (status) {
 			vu_err(vu_dev, "slave reports error: %llu\n", status);
-			return -EIO;
+			rc = -EIO;
+			goto out;
 		}
 	}
 
-	return 0;
+out:
+	spin_unlock_irqrestore(&vu_dev->sock_lock, flags);
+	return rc;
 }
 
 static int vhost_user_send_no_payload(struct virtio_uml_device *vu_dev,
@@ -324,6 +355,7 @@ static void vhost_user_reply(struct virtio_uml_device *vu_dev,
 static irqreturn_t vu_req_interrupt(int irq, void *data)
 {
 	struct virtio_uml_device *vu_dev = data;
+	struct virtqueue *vq;
 	int response = 1;
 	struct {
 		struct vhost_user_msg msg;
@@ -343,6 +375,15 @@ static irqreturn_t vu_req_interrupt(int irq, void *data)
 		virtio_config_changed(&vu_dev->vdev);
 		response = 0;
 		break;
+	case VHOST_USER_SLAVE_VRING_CALL:
+		virtio_device_for_each_vq((&vu_dev->vdev), vq) {
+			if (vq->index == msg.msg.payload.vring_state.index) {
+				response = 0;
+				vring_interrupt(0 /* ignored */, vq);
+				break;
+			}
+		}
+		break;
 	case VHOST_USER_SLAVE_IOTLB_MSG:
 		/* not supported - VIRTIO_F_IOMMU_PLATFORM */
 	case VHOST_USER_SLAVE_VRING_HOST_NOTIFIER_MSG:
@@ -684,6 +725,17 @@ static bool vu_notify(struct virtqueue *vq)
 	const uint64_t n = 1;
 	int rc;
 
+	time_travel_propagate_time();
+
+	if (info->kick_fd < 0) {
+		struct virtio_uml_device *vu_dev;
+
+		vu_dev = to_virtio_uml_device(vq->vdev);
+
+		return vhost_user_set_vring_state(vu_dev, VHOST_USER_VRING_KICK,
+						  vq->index, 0) == 0;
+	}
+
 	do {
 		rc = os_write_file(info->kick_fd, &n, sizeof(n));
 	} while (rc == -EINTR);
@@ -749,10 +801,13 @@ static void vu_del_vq(struct virtqueue *vq)
 {
 	struct virtio_uml_vq_info *info = vq->priv;
 
-	um_free_irq(VIRTIO_IRQ, vq);
+	if (info->call_fd >= 0) {
+		um_free_irq(VIRTIO_IRQ, vq);
+		os_close_file(info->call_fd);
+	}
 
-	os_close_file(info->call_fd);
-	os_close_file(info->kick_fd);
+	if (info->kick_fd >= 0)
+		os_close_file(info->kick_fd);
 
 	vring_del_virtqueue(vq);
 	kfree(info);
@@ -782,6 +837,15 @@ static int vu_setup_vq_call_fd(struct virtio_uml_device *vu_dev,
 	int call_fds[2];
 	int rc;
 
+	/* no call FD needed/desired in this case */
+	if (vu_dev->protocol_features &
+			BIT_ULL(VHOST_USER_PROTOCOL_F_INBAND_NOTIFICATIONS) &&
+	    vu_dev->protocol_features &
+			BIT_ULL(VHOST_USER_PROTOCOL_F_SLAVE_REQ)) {
+		info->call_fd = -1;
+		return 0;
+	}
+
 	/* Use a pipe for call fd, since SIGIO is not supported for eventfd */
 	rc = os_pipe(call_fds, true, true);
 	if (rc < 0)
@@ -810,6 +874,23 @@ out:
 	return rc;
 }
 
+#ifdef CONFIG_UML_TIME_TRAVEL_SUPPORT
+static void vu_defer_irq_handle(struct time_travel_event *d)
+{
+	struct virtio_uml_vq_info *info;
+
+	info = container_of(d, struct virtio_uml_vq_info, defer);
+	info->callback(info->vq);
+}
+
+static void vu_defer_irq_callback(struct virtqueue *vq)
+{
+	struct virtio_uml_vq_info *info = vq->priv;
+
+	time_travel_add_irq_event(&info->defer);
+}
+#endif
+
 static struct virtqueue *vu_setup_vq(struct virtio_device *vdev,
 				     unsigned index, vq_callback_t *callback,
 				     const char *name, bool ctx)
@@ -829,6 +910,19 @@ static struct virtqueue *vu_setup_vq(struct virtio_device *vdev,
 	snprintf(info->name, sizeof(info->name), "%s.%d-%s", pdev->name,
 		 pdev->id, name);
 
+#ifdef CONFIG_UML_TIME_TRAVEL_SUPPORT
+	/*
+	 * When we get an interrupt, we must bounce it through the simulation
+	 * calendar (the simtime device), except for the simtime device itself
+	 * since that's part of the simulation control.
+	 */
+	if (time_travel_mode == TT_MODE_EXTERNAL && callback) {
+		info->callback = callback;
+		callback = vu_defer_irq_callback;
+		time_travel_set_event_fn(&info->defer, vu_defer_irq_handle);
+	}
+#endif
+
 	vq = vring_create_virtqueue(index, num, PAGE_SIZE, vdev, true, true,
 				    ctx, vu_notify, callback, info->name);
 	if (!vq) {
@@ -837,11 +931,19 @@ static struct virtqueue *vu_setup_vq(struct virtio_device *vdev,
 	}
 	vq->priv = info;
 	num = virtqueue_get_vring_size(vq);
+#ifdef CONFIG_UML_TIME_TRAVEL_SUPPORT
+	info->vq = vq;
+#endif
 
-	rc = os_eventfd(0, 0);
-	if (rc < 0)
-		goto error_kick;
-	info->kick_fd = rc;
+	if (vu_dev->protocol_features &
+			BIT_ULL(VHOST_USER_PROTOCOL_F_INBAND_NOTIFICATIONS)) {
+		info->kick_fd = -1;
+	} else {
+		rc = os_eventfd(0, 0);
+		if (rc < 0)
+			goto error_kick;
+		info->kick_fd = rc;
+	}
 
 	rc = vu_setup_vq_call_fd(vu_dev, vq);
 	if (rc)
@@ -866,10 +968,13 @@ static struct virtqueue *vu_setup_vq(struct virtio_device *vdev,
 	return vq;
 
 error_setup:
-	um_free_irq(VIRTIO_IRQ, vq);
-	os_close_file(info->call_fd);
+	if (info->call_fd >= 0) {
+		um_free_irq(VIRTIO_IRQ, vq);
+		os_close_file(info->call_fd);
+	}
 error_call:
-	os_close_file(info->kick_fd);
+	if (info->kick_fd >= 0)
+		os_close_file(info->kick_fd);
 error_kick:
 	vring_del_virtqueue(vq);
 error_create:
@@ -908,10 +1013,12 @@ static int vu_find_vqs(struct virtio_device *vdev, unsigned nvqs,
 	list_for_each_entry(vq, &vdev->vqs, list) {
 		struct virtio_uml_vq_info *info = vq->priv;
 
-		rc = vhost_user_set_vring_kick(vu_dev, vq->index,
-					       info->kick_fd);
-		if (rc)
-			goto error_setup;
+		if (info->kick_fd >= 0) {
+			rc = vhost_user_set_vring_kick(vu_dev, vq->index,
+						       info->kick_fd);
+			if (rc)
+				goto error_setup;
+		}
 
 		rc = vhost_user_set_vring_enable(vu_dev, vq->index, true);
 		if (rc)
@@ -1008,6 +1115,8 @@ static int virtio_uml_probe(struct platform_device *pdev)
 		return rc;
 	vu_dev->sock = rc;
 
+	spin_lock_init(&vu_dev->sock_lock);
+
 	rc = vhost_user_init(vu_dev);
 	if (rc)
 		goto error_init;
diff --git a/arch/um/include/asm/Kbuild b/arch/um/include/asm/Kbuild
index db7d9d4e30d8..8d435f8a6dec 100644
--- a/arch/um/include/asm/Kbuild
+++ b/arch/um/include/asm/Kbuild
@@ -3,7 +3,6 @@ generic-y += bpf_perf_event.h
 generic-y += bug.h
 generic-y += compat.h
 generic-y += current.h
-generic-y += delay.h
 generic-y += device.h
 generic-y += emergency-restart.h
 generic-y += exec.h
diff --git a/arch/um/include/asm/delay.h b/arch/um/include/asm/delay.h
new file mode 100644
index 000000000000..56fc2b8f2dd0
--- /dev/null
+++ b/arch/um/include/asm/delay.h
@@ -0,0 +1,30 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+#ifndef __UM_DELAY_H
+#define __UM_DELAY_H
+#include <asm-generic/delay.h>
+#include <linux/time-internal.h>
+
+static inline void um_ndelay(unsigned long nsecs)
+{
+	if (time_travel_mode == TT_MODE_INFCPU ||
+	    time_travel_mode == TT_MODE_EXTERNAL) {
+		time_travel_ndelay(nsecs);
+		return;
+	}
+	ndelay(nsecs);
+}
+#undef ndelay
+#define ndelay um_ndelay
+
+static inline void um_udelay(unsigned long usecs)
+{
+	if (time_travel_mode == TT_MODE_INFCPU ||
+	    time_travel_mode == TT_MODE_EXTERNAL) {
+		time_travel_ndelay(1000 * usecs);
+		return;
+	}
+	udelay(usecs);
+}
+#undef udelay
+#define udelay um_udelay
+#endif /* __UM_DELAY_H */
diff --git a/arch/um/include/linux/time-internal.h b/arch/um/include/linux/time-internal.h
new file mode 100644
index 000000000000..f3b03d39a854
--- /dev/null
+++ b/arch/um/include/linux/time-internal.h
@@ -0,0 +1,84 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Copyright (C) 2012 - 2014 Cisco Systems
+ * Copyright (C) 2000 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
+ */
+
+#ifndef __TIMER_INTERNAL_H__
+#define __TIMER_INTERNAL_H__
+#include <linux/list.h>
+
+#define TIMER_MULTIPLIER 256
+#define TIMER_MIN_DELTA  500
+
+enum time_travel_mode {
+	TT_MODE_OFF,
+	TT_MODE_BASIC,
+	TT_MODE_INFCPU,
+	TT_MODE_EXTERNAL,
+};
+
+#ifdef CONFIG_UML_TIME_TRAVEL_SUPPORT
+struct time_travel_event {
+	unsigned long long time;
+	void (*fn)(struct time_travel_event *d);
+	struct list_head list;
+	bool pending, onstack;
+};
+
+extern enum time_travel_mode time_travel_mode;
+
+void time_travel_sleep(unsigned long long duration);
+
+static inline void
+time_travel_set_event_fn(struct time_travel_event *e,
+			 void (*fn)(struct time_travel_event *d))
+{
+	e->fn = fn;
+}
+
+void __time_travel_propagate_time(void);
+
+static inline void time_travel_propagate_time(void)
+{
+	if (time_travel_mode == TT_MODE_EXTERNAL)
+		__time_travel_propagate_time();
+}
+
+void __time_travel_wait_readable(int fd);
+
+static inline void time_travel_wait_readable(int fd)
+{
+	if (time_travel_mode == TT_MODE_EXTERNAL)
+		__time_travel_wait_readable(fd);
+}
+
+void time_travel_add_irq_event(struct time_travel_event *e);
+#else
+struct time_travel_event {
+};
+
+#define time_travel_mode TT_MODE_OFF
+
+static inline void time_travel_sleep(unsigned long long duration)
+{
+}
+
+/* this is a macro so the event/function need not exist */
+#define time_travel_set_event_fn(e, fn) do {} while (0)
+
+static inline void time_travel_propagate_time(void)
+{
+}
+
+static inline void time_travel_wait_readable(int fd)
+{
+}
+#endif /* CONFIG_UML_TIME_TRAVEL_SUPPORT */
+
+/*
+ * Without CONFIG_UML_TIME_TRAVEL_SUPPORT this is a linker error if used,
+ * which is intentional since we really shouldn't link it in that case.
+ */
+void time_travel_ndelay(unsigned long nsec);
+#endif /* __TIMER_INTERNAL_H__ */
diff --git a/arch/um/include/shared/os.h b/arch/um/include/shared/os.h
index 0f30204b6afa..f467d28fc0b4 100644
--- a/arch/um/include/shared/os.h
+++ b/arch/um/include/shared/os.h
@@ -181,6 +181,7 @@ extern int os_falloc_punch(int fd, unsigned long long offset, int count);
 extern int os_eventfd(unsigned int initval, int flags);
 extern int os_sendmsg_fds(int fd, const void *buf, unsigned int len,
 			  const int *fds, unsigned int fds_num);
+int os_poll(unsigned int n, const int *fds);
 
 /* start_up.c */
 extern void os_early_checks(void);
diff --git a/arch/um/include/shared/timer-internal.h b/arch/um/include/shared/timer-internal.h
deleted file mode 100644
index 2d2d13c9b46f..000000000000
--- a/arch/um/include/shared/timer-internal.h
+++ /dev/null
@@ -1,76 +0,0 @@
-/* SPDX-License-Identifier: GPL-2.0 */
-/*
- * Copyright (C) 2012 - 2014 Cisco Systems
- * Copyright (C) 2000 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
- */
-
-#ifndef __TIMER_INTERNAL_H__
-#define __TIMER_INTERNAL_H__
-
-#define TIMER_MULTIPLIER 256
-#define TIMER_MIN_DELTA  500
-
-enum time_travel_mode {
-	TT_MODE_OFF,
-	TT_MODE_BASIC,
-	TT_MODE_INFCPU,
-};
-
-enum time_travel_timer_mode {
-	TT_TMR_DISABLED,
-	TT_TMR_ONESHOT,
-	TT_TMR_PERIODIC,
-};
-
-#ifdef CONFIG_UML_TIME_TRAVEL_SUPPORT
-extern enum time_travel_mode time_travel_mode;
-extern unsigned long long time_travel_time;
-extern enum time_travel_timer_mode time_travel_timer_mode;
-extern unsigned long long time_travel_timer_expiry;
-extern unsigned long long time_travel_timer_interval;
-
-static inline void time_travel_set_time(unsigned long long ns)
-{
-	time_travel_time = ns;
-}
-
-static inline void time_travel_set_timer_mode(enum time_travel_timer_mode mode)
-{
-	time_travel_timer_mode = mode;
-}
-
-static inline void time_travel_set_timer_expiry(unsigned long long expiry)
-{
-	time_travel_timer_expiry = expiry;
-}
-
-static inline void time_travel_set_timer_interval(unsigned long long interval)
-{
-	time_travel_timer_interval = interval;
-}
-#else
-#define time_travel_mode TT_MODE_OFF
-#define time_travel_time 0
-#define time_travel_timer_expiry 0
-#define time_travel_timer_interval 0
-
-static inline void time_travel_set_time(unsigned long long ns)
-{
-}
-
-static inline void time_travel_set_timer_mode(enum time_travel_timer_mode mode)
-{
-}
-
-static inline void time_travel_set_timer_expiry(unsigned long long expiry)
-{
-}
-
-static inline void time_travel_set_timer_interval(unsigned long long interval)
-{
-}
-
-#define time_travel_timer_mode TT_TMR_DISABLED
-#endif
-
-#endif
diff --git a/arch/um/kernel/kmsg_dump.c b/arch/um/kernel/kmsg_dump.c
index 98bdf69e4c2e..e4abac6c9727 100644
--- a/arch/um/kernel/kmsg_dump.c
+++ b/arch/um/kernel/kmsg_dump.c
@@ -9,20 +9,19 @@ static void kmsg_dumper_stdout(struct kmsg_dumper *dumper,
 				enum kmsg_dump_reason reason)
 {
 	static char line[1024];
-
+	struct console *con;
 	size_t len = 0;
-	bool con_available = false;
 
 	/* only dump kmsg when no console is available */
 	if (!console_trylock())
 		return;
 
-	if (console_drivers != NULL)
-		con_available = true;
+	for_each_console(con)
+		break;
 
 	console_unlock();
 
-	if (con_available == true)
+	if (con)
 		return;
 
 	printf("kmsg_dump:\n");
diff --git a/arch/um/kernel/process.c b/arch/um/kernel/process.c
index 56a094182bf5..cbe33af2a880 100644
--- a/arch/um/kernel/process.c
+++ b/arch/um/kernel/process.c
@@ -32,7 +32,7 @@
 #include <kern_util.h>
 #include <os.h>
 #include <skas.h>
-#include <timer-internal.h>
+#include <linux/time-internal.h>
 
 /*
  * This is a per-cpu array.  A processor only modifies its entry and it only
@@ -203,43 +203,6 @@ void initial_thread_cb(void (*proc)(void *), void *arg)
 	kmalloc_ok = save_kmalloc_ok;
 }
 
-static void time_travel_sleep(unsigned long long duration)
-{
-	unsigned long long next = time_travel_time + duration;
-
-	if (time_travel_mode != TT_MODE_INFCPU)
-		os_timer_disable();
-
-	while (time_travel_timer_mode == TT_TMR_PERIODIC &&
-	       time_travel_timer_expiry < time_travel_time)
-		time_travel_set_timer_expiry(time_travel_timer_expiry +
-					     time_travel_timer_interval);
-
-	if (time_travel_timer_mode != TT_TMR_DISABLED &&
-	    time_travel_timer_expiry < next) {
-		if (time_travel_timer_mode == TT_TMR_ONESHOT)
-			time_travel_set_timer_mode(TT_TMR_DISABLED);
-		/*
-		 * In basic mode, time_travel_time will be adjusted in
-		 * the timer IRQ handler so it works even when the signal
-		 * comes from the OS timer, see there.
-		 */
-		if (time_travel_mode != TT_MODE_BASIC)
-			time_travel_set_time(time_travel_timer_expiry);
-
-		deliver_alarm();
-	} else {
-		time_travel_set_time(next);
-	}
-
-	if (time_travel_mode != TT_MODE_INFCPU) {
-		if (time_travel_timer_mode == TT_TMR_PERIODIC)
-			os_timer_set_interval(time_travel_timer_interval);
-		else if (time_travel_timer_mode == TT_TMR_ONESHOT)
-			os_timer_one_shot(time_travel_timer_expiry - next);
-	}
-}
-
 static void um_idle_sleep(void)
 {
 	unsigned long long duration = UM_NSEC_PER_SEC;
diff --git a/arch/um/kernel/skas/syscall.c b/arch/um/kernel/skas/syscall.c
index 40d90dddf3f1..0a12d5a09217 100644
--- a/arch/um/kernel/skas/syscall.c
+++ b/arch/um/kernel/skas/syscall.c
@@ -10,7 +10,7 @@
 #include <sysdep/ptrace.h>
 #include <sysdep/ptrace_user.h>
 #include <sysdep/syscalls.h>
-#include <shared/timer-internal.h>
+#include <linux/time-internal.h>
 
 void handle_syscall(struct uml_pt_regs *r)
 {
@@ -24,7 +24,8 @@ void handle_syscall(struct uml_pt_regs *r)
 	 * went to sleep, even if said userspace interacts with the kernel in
 	 * various ways.
 	 */
-	if (time_travel_mode == TT_MODE_INFCPU)
+	if (time_travel_mode == TT_MODE_INFCPU ||
+	    time_travel_mode == TT_MODE_EXTERNAL)
 		schedule();
 
 	/* Initialize the syscall number and default return value. */
diff --git a/arch/um/kernel/time.c b/arch/um/kernel/time.c
index 94ea87bd231c..25eaa6a0c658 100644
--- a/arch/um/kernel/time.c
+++ b/arch/um/kernel/time.c
@@ -4,6 +4,7 @@
  * Copyright (C) 2015 Thomas Meyer (thomas@m3y3r.de)
  * Copyright (C) 2012-2014 Cisco Systems
  * Copyright (C) 2000 - 2007 Jeff Dike (jdike@{addtoit,linux.intel}.com)
+ * Copyright (C) 2019 Intel Corporation
  */
 
 #include <linux/clockchips.h>
@@ -18,21 +19,484 @@
 #include <asm/param.h>
 #include <kern_util.h>
 #include <os.h>
-#include <timer-internal.h>
+#include <linux/time-internal.h>
+#include <linux/um_timetravel.h>
 #include <shared/init.h>
 
 #ifdef CONFIG_UML_TIME_TRAVEL_SUPPORT
 enum time_travel_mode time_travel_mode;
-unsigned long long time_travel_time;
-enum time_travel_timer_mode time_travel_timer_mode;
-unsigned long long time_travel_timer_expiry;
-unsigned long long time_travel_timer_interval;
+EXPORT_SYMBOL_GPL(time_travel_mode);
 
 static bool time_travel_start_set;
 static unsigned long long time_travel_start;
-#else
+static unsigned long long time_travel_time;
+static LIST_HEAD(time_travel_events);
+static unsigned long long time_travel_timer_interval;
+static unsigned long long time_travel_next_event;
+static struct time_travel_event time_travel_timer_event;
+static int time_travel_ext_fd = -1;
+static unsigned int time_travel_ext_waiting;
+static bool time_travel_ext_prev_request_valid;
+static unsigned long long time_travel_ext_prev_request;
+static bool time_travel_ext_free_until_valid;
+static unsigned long long time_travel_ext_free_until;
+
+static void time_travel_set_time(unsigned long long ns)
+{
+	if (unlikely(ns < time_travel_time))
+		panic("time-travel: time goes backwards %lld -> %lld\n",
+		      time_travel_time, ns);
+	time_travel_time = ns;
+}
+
+enum time_travel_message_handling {
+	TTMH_IDLE,
+	TTMH_POLL,
+	TTMH_READ,
+};
+
+static void time_travel_handle_message(struct um_timetravel_msg *msg,
+				       enum time_travel_message_handling mode)
+{
+	struct um_timetravel_msg resp = {
+		.op = UM_TIMETRAVEL_ACK,
+	};
+	int ret;
+
+	/*
+	 * Poll outside the locked section (if we're not called to only read
+	 * the response) so we can get interrupts for e.g. virtio while we're
+	 * here, but then we need to lock to not get interrupted between the
+	 * read of the message and write of the ACK.
+	 */
+	if (mode != TTMH_READ) {
+		while (os_poll(1, &time_travel_ext_fd) != 0) {
+			if (mode == TTMH_IDLE) {
+				BUG_ON(!irqs_disabled());
+				local_irq_enable();
+				local_irq_disable();
+			}
+		}
+	}
+
+	ret = os_read_file(time_travel_ext_fd, msg, sizeof(*msg));
+
+	if (ret == 0)
+		panic("time-travel external link is broken\n");
+	if (ret != sizeof(*msg))
+		panic("invalid time-travel message - %d bytes\n", ret);
+
+	switch (msg->op) {
+	default:
+		WARN_ONCE(1, "time-travel: unexpected message %lld\n",
+			  (unsigned long long)msg->op);
+		break;
+	case UM_TIMETRAVEL_ACK:
+		return;
+	case UM_TIMETRAVEL_RUN:
+		time_travel_set_time(msg->time);
+		break;
+	case UM_TIMETRAVEL_FREE_UNTIL:
+		time_travel_ext_free_until_valid = true;
+		time_travel_ext_free_until = msg->time;
+		break;
+	}
+
+	os_write_file(time_travel_ext_fd, &resp, sizeof(resp));
+}
+
+static u64 time_travel_ext_req(u32 op, u64 time)
+{
+	static int seq;
+	int mseq = ++seq;
+	struct um_timetravel_msg msg = {
+		.op = op,
+		.time = time,
+		.seq = mseq,
+	};
+	unsigned long flags;
+
+	/*
+	 * We need to save interrupts here and only restore when we
+	 * got the ACK - otherwise we can get interrupted and send
+	 * another request while we're still waiting for an ACK, but
+	 * the peer doesn't know we got interrupted and will send
+	 * the ACKs in the same order as the message, but we'd need
+	 * to see them in the opposite order ...
+	 *
+	 * This wouldn't matter *too* much, but some ACKs carry the
+	 * current time (for UM_TIMETRAVEL_GET) and getting another
+	 * ACK without a time would confuse us a lot!
+	 *
+	 * The sequence number assignment that happens here lets us
+	 * debug such message handling issues more easily.
+	 */
+	local_irq_save(flags);
+	os_write_file(time_travel_ext_fd, &msg, sizeof(msg));
+
+	while (msg.op != UM_TIMETRAVEL_ACK)
+		time_travel_handle_message(&msg, TTMH_READ);
+
+	if (msg.seq != mseq)
+		panic("time-travel: ACK message has different seqno! op=%d, seq=%d != %d time=%lld\n",
+		      msg.op, msg.seq, mseq, msg.time);
+
+	if (op == UM_TIMETRAVEL_GET)
+		time_travel_set_time(msg.time);
+	local_irq_restore(flags);
+
+	return msg.time;
+}
+
+void __time_travel_wait_readable(int fd)
+{
+	int fds[2] = { fd, time_travel_ext_fd };
+	int ret;
+
+	if (time_travel_mode != TT_MODE_EXTERNAL)
+		return;
+
+	while ((ret = os_poll(2, fds))) {
+		struct um_timetravel_msg msg;
+
+		if (ret == 1)
+			time_travel_handle_message(&msg, TTMH_READ);
+	}
+}
+EXPORT_SYMBOL_GPL(__time_travel_wait_readable);
+
+static void time_travel_ext_update_request(unsigned long long time)
+{
+	if (time_travel_mode != TT_MODE_EXTERNAL)
+		return;
+
+	/* asked for exactly this time previously */
+	if (time_travel_ext_prev_request_valid &&
+	    time == time_travel_ext_prev_request)
+		return;
+
+	time_travel_ext_prev_request = time;
+	time_travel_ext_prev_request_valid = true;
+	time_travel_ext_req(UM_TIMETRAVEL_REQUEST, time);
+}
+
+void __time_travel_propagate_time(void)
+{
+	time_travel_ext_req(UM_TIMETRAVEL_UPDATE, time_travel_time);
+}
+EXPORT_SYMBOL_GPL(__time_travel_propagate_time);
+
+/* returns true if we must do a wait to the simtime device */
+static bool time_travel_ext_request(unsigned long long time)
+{
+	/*
+	 * If we received an external sync point ("free until") then we
+	 * don't have to request/wait for anything until then, unless
+	 * we're already waiting.
+	 */
+	if (!time_travel_ext_waiting && time_travel_ext_free_until_valid &&
+	    time < time_travel_ext_free_until)
+		return false;
+
+	time_travel_ext_update_request(time);
+	return true;
+}
+
+static void time_travel_ext_wait(bool idle)
+{
+	struct um_timetravel_msg msg = {
+		.op = UM_TIMETRAVEL_ACK,
+	};
+
+	time_travel_ext_prev_request_valid = false;
+	time_travel_ext_waiting++;
+
+	time_travel_ext_req(UM_TIMETRAVEL_WAIT, -1);
+
+	/*
+	 * Here we are deep in the idle loop, so we have to break out of the
+	 * kernel abstraction in a sense and implement this in terms of the
+	 * UML system waiting on the VQ interrupt while sleeping, when we get
+	 * the signal it'll call time_travel_ext_vq_notify_done() completing the
+	 * call.
+	 */
+	while (msg.op != UM_TIMETRAVEL_RUN)
+		time_travel_handle_message(&msg, idle ? TTMH_IDLE : TTMH_POLL);
+
+	time_travel_ext_waiting--;
+
+	/* we might request more stuff while polling - reset when we run */
+	time_travel_ext_prev_request_valid = false;
+}
+
+static void time_travel_ext_get_time(void)
+{
+	time_travel_ext_req(UM_TIMETRAVEL_GET, -1);
+}
+
+static void __time_travel_update_time(unsigned long long ns, bool idle)
+{
+	if (time_travel_mode == TT_MODE_EXTERNAL && time_travel_ext_request(ns))
+		time_travel_ext_wait(idle);
+	else
+		time_travel_set_time(ns);
+}
+
+static struct time_travel_event *time_travel_first_event(void)
+{
+	return list_first_entry_or_null(&time_travel_events,
+					struct time_travel_event,
+					list);
+}
+
+static void __time_travel_add_event(struct time_travel_event *e,
+				    unsigned long long time)
+{
+	struct time_travel_event *tmp;
+	bool inserted = false;
+
+	if (WARN(time_travel_mode == TT_MODE_BASIC &&
+		 e != &time_travel_timer_event,
+		 "only timer events can be handled in basic mode"))
+		return;
+
+	if (e->pending)
+		return;
+
+	e->pending = true;
+	e->time = time;
+
+	list_for_each_entry(tmp, &time_travel_events, list) {
+		/*
+		 * Add the new entry before one with higher time,
+		 * or if they're equal and both on stack, because
+		 * in that case we need to unwind the stack in the
+		 * right order, and the later event (timer sleep
+		 * or such) must be dequeued first.
+		 */
+		if ((tmp->time > e->time) ||
+		    (tmp->time == e->time && tmp->onstack && e->onstack)) {
+			list_add_tail(&e->list, &tmp->list);
+			inserted = true;
+			break;
+		}
+	}
+
+	if (!inserted)
+		list_add_tail(&e->list, &time_travel_events);
+
+	tmp = time_travel_first_event();
+	time_travel_ext_update_request(tmp->time);
+	time_travel_next_event = tmp->time;
+}
+
+static void time_travel_add_event(struct time_travel_event *e,
+				  unsigned long long time)
+{
+	if (WARN_ON(!e->fn))
+		return;
+
+	__time_travel_add_event(e, time);
+}
+
+void time_travel_periodic_timer(struct time_travel_event *e)
+{
+	time_travel_add_event(&time_travel_timer_event,
+			      time_travel_time + time_travel_timer_interval);
+	deliver_alarm();
+}
+
+static void time_travel_deliver_event(struct time_travel_event *e)
+{
+	if (e == &time_travel_timer_event) {
+		/*
+		 * deliver_alarm() does the irq_enter/irq_exit
+		 * by itself, so must handle it specially here
+		 */
+		e->fn(e);
+	} else {
+		unsigned long flags;
+
+		local_irq_save(flags);
+		irq_enter();
+		e->fn(e);
+		irq_exit();
+		local_irq_restore(flags);
+	}
+}
+
+static bool time_travel_del_event(struct time_travel_event *e)
+{
+	if (!e->pending)
+		return false;
+	list_del(&e->list);
+	e->pending = false;
+	return true;
+}
+
+static void time_travel_update_time(unsigned long long next, bool idle)
+{
+	struct time_travel_event ne = {
+		.onstack = true,
+	};
+	struct time_travel_event *e;
+	bool finished = idle;
+
+	/* add it without a handler - we deal with that specifically below */
+	__time_travel_add_event(&ne, next);
+
+	do {
+		e = time_travel_first_event();
+
+		BUG_ON(!e);
+		__time_travel_update_time(e->time, idle);
+
+		/* new events may have been inserted while we were waiting */
+		if (e == time_travel_first_event()) {
+			BUG_ON(!time_travel_del_event(e));
+			BUG_ON(time_travel_time != e->time);
+
+			if (e == &ne) {
+				finished = true;
+			} else {
+				if (e->onstack)
+					panic("On-stack event dequeued outside of the stack! time=%lld, event time=%lld, event=%pS\n",
+					      time_travel_time, e->time, e);
+				time_travel_deliver_event(e);
+			}
+		}
+
+		e = time_travel_first_event();
+		if (e)
+			time_travel_ext_update_request(e->time);
+	} while (ne.pending && !finished);
+
+	time_travel_del_event(&ne);
+}
+
+void time_travel_ndelay(unsigned long nsec)
+{
+	time_travel_update_time(time_travel_time + nsec, false);
+}
+EXPORT_SYMBOL(time_travel_ndelay);
+
+void time_travel_add_irq_event(struct time_travel_event *e)
+{
+	BUG_ON(time_travel_mode != TT_MODE_EXTERNAL);
+
+	time_travel_ext_get_time();
+	/*
+	 * We could model interrupt latency here, for now just
+	 * don't have any latency at all and request the exact
+	 * same time (again) to run the interrupt...
+	 */
+	time_travel_add_event(e, time_travel_time);
+}
+EXPORT_SYMBOL_GPL(time_travel_add_irq_event);
+
+static void time_travel_oneshot_timer(struct time_travel_event *e)
+{
+	deliver_alarm();
+}
+
+void time_travel_sleep(unsigned long long duration)
+{
+	unsigned long long next = time_travel_time + duration;
+
+	if (time_travel_mode == TT_MODE_BASIC)
+		os_timer_disable();
+
+	time_travel_update_time(next, true);
+
+	if (time_travel_mode == TT_MODE_BASIC &&
+	    time_travel_timer_event.pending) {
+		if (time_travel_timer_event.fn == time_travel_periodic_timer) {
+			/*
+			 * This is somewhat wrong - we should get the first
+			 * one sooner like the os_timer_one_shot() below...
+			 */
+			os_timer_set_interval(time_travel_timer_interval);
+		} else {
+			os_timer_one_shot(time_travel_timer_event.time - next);
+		}
+	}
+}
+
+static void time_travel_handle_real_alarm(void)
+{
+	time_travel_set_time(time_travel_next_event);
+
+	time_travel_del_event(&time_travel_timer_event);
+
+	if (time_travel_timer_event.fn == time_travel_periodic_timer)
+		time_travel_add_event(&time_travel_timer_event,
+				      time_travel_time +
+				      time_travel_timer_interval);
+}
+
+static void time_travel_set_interval(unsigned long long interval)
+{
+	time_travel_timer_interval = interval;
+}
+
+static int time_travel_connect_external(const char *socket)
+{
+	const char *sep;
+	unsigned long long id = (unsigned long long)-1;
+	int rc;
+
+	if ((sep = strchr(socket, ':'))) {
+		char buf[25] = {};
+		if (sep - socket > sizeof(buf) - 1)
+			goto invalid_number;
+
+		memcpy(buf, socket, sep - socket);
+		if (kstrtoull(buf, 0, &id)) {
+invalid_number:
+			panic("time-travel: invalid external ID in string '%s'\n",
+			      socket);
+			return -EINVAL;
+		}
+
+		socket = sep + 1;
+	}
+
+	rc = os_connect_socket(socket);
+	if (rc < 0) {
+		panic("time-travel: failed to connect to external socket %s\n",
+		      socket);
+		return rc;
+	}
+
+	time_travel_ext_fd = rc;
+
+	time_travel_ext_req(UM_TIMETRAVEL_START, id);
+
+	return 1;
+}
+#else /* CONFIG_UML_TIME_TRAVEL_SUPPORT */
 #define time_travel_start_set 0
 #define time_travel_start 0
+#define time_travel_time 0
+
+static inline void time_travel_update_time(unsigned long long ns, bool retearly)
+{
+}
+
+static inline void time_travel_handle_real_alarm(void)
+{
+}
+
+static void time_travel_set_interval(unsigned long long interval)
+{
+}
+
+/* fail link if this actually gets used */
+extern u64 time_travel_ext_req(u32 op, u64 time);
+
+/* these are empty macros so the struct/fn need not exist */
+#define time_travel_add_event(e, time) do { } while (0)
+#define time_travel_del_event(e) do { } while (0)
 #endif
 
 void timer_handler(int sig, struct siginfo *unused_si, struct uml_pt_regs *regs)
@@ -48,7 +512,7 @@ void timer_handler(int sig, struct siginfo *unused_si, struct uml_pt_regs *regs)
 	 * never get any real signals from the OS.
 	 */
 	if (time_travel_mode == TT_MODE_BASIC)
-		time_travel_set_time(time_travel_timer_expiry);
+		time_travel_handle_real_alarm();
 
 	local_irq_save(flags);
 	do_IRQ(TIMER_IRQ, regs);
@@ -58,9 +522,10 @@ void timer_handler(int sig, struct siginfo *unused_si, struct uml_pt_regs *regs)
 static int itimer_shutdown(struct clock_event_device *evt)
 {
 	if (time_travel_mode != TT_MODE_OFF)
-		time_travel_set_timer_mode(TT_TMR_DISABLED);
+		time_travel_del_event(&time_travel_timer_event);
 
-	if (time_travel_mode != TT_MODE_INFCPU)
+	if (time_travel_mode != TT_MODE_INFCPU &&
+	    time_travel_mode != TT_MODE_EXTERNAL)
 		os_timer_disable();
 
 	return 0;
@@ -71,12 +536,16 @@ static int itimer_set_periodic(struct clock_event_device *evt)
 	unsigned long long interval = NSEC_PER_SEC / HZ;
 
 	if (time_travel_mode != TT_MODE_OFF) {
-		time_travel_set_timer_mode(TT_TMR_PERIODIC);
-		time_travel_set_timer_expiry(time_travel_time + interval);
-		time_travel_set_timer_interval(interval);
+		time_travel_del_event(&time_travel_timer_event);
+		time_travel_set_event_fn(&time_travel_timer_event,
+					 time_travel_periodic_timer);
+		time_travel_set_interval(interval);
+		time_travel_add_event(&time_travel_timer_event,
+				      time_travel_time + interval);
 	}
 
-	if (time_travel_mode != TT_MODE_INFCPU)
+	if (time_travel_mode != TT_MODE_INFCPU &&
+	    time_travel_mode != TT_MODE_EXTERNAL)
 		os_timer_set_interval(interval);
 
 	return 0;
@@ -88,11 +557,15 @@ static int itimer_next_event(unsigned long delta,
 	delta += 1;
 
 	if (time_travel_mode != TT_MODE_OFF) {
-		time_travel_set_timer_mode(TT_TMR_ONESHOT);
-		time_travel_set_timer_expiry(time_travel_time + delta);
+		time_travel_del_event(&time_travel_timer_event);
+		time_travel_set_event_fn(&time_travel_timer_event,
+					 time_travel_oneshot_timer);
+		time_travel_add_event(&time_travel_timer_event,
+				      time_travel_time + delta);
 	}
 
-	if (time_travel_mode != TT_MODE_INFCPU)
+	if (time_travel_mode != TT_MODE_INFCPU &&
+	    time_travel_mode != TT_MODE_EXTERNAL)
 		return os_timer_one_shot(delta);
 
 	return 0;
@@ -143,8 +616,17 @@ static u64 timer_read(struct clocksource *cs)
 		 * stuck in loops that expect time to move more than the
 		 * exact requested sleep amount, e.g. python's socket server,
 		 * see https://bugs.python.org/issue37026.
+		 *
+		 * However, don't do that when we're in interrupt or such as
+		 * then we might recurse into our own processing, and get to
+		 * even more waiting, and that's not good - it messes up the
+		 * "what do I do next" and onstack event we use to know when
+		 * to return from time_travel_update_time().
 		 */
-		time_travel_set_time(time_travel_time + TIMER_MULTIPLIER);
+		if (!irqs_disabled() && !in_interrupt() && !in_softirq())
+			time_travel_update_time(time_travel_time +
+						TIMER_MULTIPLIER,
+						false);
 		return time_travel_time / TIMER_MULTIPLIER;
 	}
 
@@ -188,6 +670,8 @@ void read_persistent_clock64(struct timespec64 *ts)
 
 	if (time_travel_start_set)
 		nsecs = time_travel_start + time_travel_time;
+	else if (time_travel_mode == TT_MODE_EXTERNAL)
+		nsecs = time_travel_ext_req(UM_TIMETRAVEL_GET_TOD, -1);
 	else
 		nsecs = os_persistent_clock_emulation();
 
@@ -204,7 +688,8 @@ void __init time_init(void)
 #ifdef CONFIG_UML_TIME_TRAVEL_SUPPORT
 unsigned long calibrate_delay_is_known(void)
 {
-	if (time_travel_mode == TT_MODE_INFCPU)
+	if (time_travel_mode == TT_MODE_INFCPU ||
+	    time_travel_mode == TT_MODE_EXTERNAL)
 		return 1;
 	return 0;
 }
@@ -218,6 +703,13 @@ int setup_time_travel(char *str)
 		return 1;
 	}
 
+	if (strncmp(str, "=ext:", 5) == 0) {
+		time_travel_mode = TT_MODE_EXTERNAL;
+		timer_clockevent.name = "time-travel-timer-external";
+		timer_clocksource.name = "time-travel-clock-external";
+		return time_travel_connect_external(str + 5);
+	}
+
 	if (!*str) {
 		time_travel_mode = TT_MODE_BASIC;
 		timer_clockevent.name = "time-travel-timer";
@@ -242,7 +734,15 @@ __uml_help(setup_time_travel,
 "are no wall clock timers, and any CPU processing happens - as seen from the\n"
 "guest - instantly. This can be useful for accurate simulation regardless of\n"
 "debug overhead, physical CPU speed, etc. but is somewhat dangerous as it can\n"
-"easily lead to getting stuck (e.g. if anything in the system busy loops).\n");
+"easily lead to getting stuck (e.g. if anything in the system busy loops).\n"
+"\n"
+"time-travel=ext:[ID:]/path/to/socket\n"
+"This enables time travel mode similar to =inf-cpu, except the system will\n"
+"use the given socket to coordinate with a central scheduler, in order to\n"
+"have more than one system simultaneously be on simulated time. The virtio\n"
+"driver code in UML knows about this so you can also simulate networks and\n"
+"devices using it, assuming the device has the right capabilities.\n"
+"The optional ID is a 64-bit integer that's sent to the central scheduler.\n");
 
 int setup_time_travel_start(char *str)
 {
diff --git a/arch/um/kernel/uml.lds.S b/arch/um/kernel/uml.lds.S
index 9f21443be2c9..3b6dab3d4501 100644
--- a/arch/um/kernel/uml.lds.S
+++ b/arch/um/kernel/uml.lds.S
@@ -19,10 +19,10 @@ SECTIONS
   __binary_start = START;
 
   . = START + SIZEOF_HEADERS;
+  . = ALIGN(PAGE_SIZE);
 
   _text = .;
   INIT_TEXT_SECTION(0)
-  . = ALIGN(PAGE_SIZE);
 
   .text      :
   {
diff --git a/arch/um/os-Linux/file.c b/arch/um/os-Linux/file.c
index fbda10535dab..26ecbd64c409 100644
--- a/arch/um/os-Linux/file.c
+++ b/arch/um/os-Linux/file.c
@@ -5,9 +5,11 @@
 
 #include <stdio.h>
 #include <unistd.h>
+#include <stdlib.h>
 #include <errno.h>
 #include <fcntl.h>
 #include <signal.h>
+#include <linux/falloc.h>
 #include <sys/ioctl.h>
 #include <sys/mount.h>
 #include <sys/socket.h>
@@ -16,6 +18,7 @@
 #include <sys/un.h>
 #include <sys/types.h>
 #include <sys/eventfd.h>
+#include <poll.h>
 #include <os.h>
 
 static void copy_stat(struct uml_stat *dst, const struct stat64 *src)
@@ -664,3 +667,31 @@ int os_sendmsg_fds(int fd, const void *buf, unsigned int len, const int *fds,
 		return -errno;
 	return err;
 }
+
+int os_poll(unsigned int n, const int *fds)
+{
+	/* currently need 2 FDs at most so avoid dynamic allocation */
+	struct pollfd pollfds[2] = {};
+	unsigned int i;
+	int ret;
+
+	if (n > ARRAY_SIZE(pollfds))
+		return -EINVAL;
+
+	for (i = 0; i < n; i++) {
+		pollfds[i].fd = fds[i];
+		pollfds[i].events = POLLIN;
+	}
+
+	ret = poll(pollfds, n, -1);
+	if (ret < 0)
+		return -errno;
+
+	/* Return the index of the available FD */
+	for (i = 0; i < n; i++) {
+		if (pollfds[i].revents)
+			return i;
+	}
+
+	return -EIO;
+}
diff --git a/arch/um/os-Linux/time.c b/arch/um/os-Linux/time.c
index 432f8e1f55c2..90f6de224c70 100644
--- a/arch/um/os-Linux/time.c
+++ b/arch/um/os-Linux/time.c
@@ -14,7 +14,6 @@
 #include <kern_util.h>
 #include <os.h>
 #include <string.h>
-#include <timer-internal.h>
 
 static timer_t event_high_res_timer = 0;
 
diff --git a/arch/um/os-Linux/umid.c b/arch/um/os-Linux/umid.c
index 44def53a11cd..9e16078a4bf8 100644
--- a/arch/um/os-Linux/umid.c
+++ b/arch/um/os-Linux/umid.c
@@ -220,11 +220,12 @@ static void __init create_pid_file(void)
 	char pid[sizeof("nnnnn\0")], *file;
 	int fd, n;
 
-	file = malloc(strlen(uml_dir) + UMID_LEN + sizeof("/pid\0"));
+	n = strlen(uml_dir) + UMID_LEN + sizeof("/pid\0");
+	file = malloc(n);
 	if (!file)
 		return;
 
-	if (umid_file_name("pid", file, sizeof(file)))
+	if (umid_file_name("pid", file, n))
 		goto out;
 
 	fd = open(file, O_RDWR | O_CREAT | O_EXCL, 0644);
diff --git a/arch/x86/um/asm/processor.h b/arch/x86/um/asm/processor.h
index 593d5f3902bd..478710384b34 100644
--- a/arch/x86/um/asm/processor.h
+++ b/arch/x86/um/asm/processor.h
@@ -1,6 +1,7 @@
 /* SPDX-License-Identifier: GPL-2.0 */
 #ifndef __UM_PROCESSOR_H
 #define __UM_PROCESSOR_H
+#include <linux/time-internal.h>
 
 /* include faultinfo structure */
 #include <sysdep/faultinfo.h>
@@ -21,12 +22,19 @@
 #include <asm/user.h>
 
 /* REP NOP (PAUSE) is a good thing to insert into busy-wait loops. */
-static inline void rep_nop(void)
+static __always_inline void rep_nop(void)
 {
 	__asm__ __volatile__("rep;nop": : :"memory");
 }
 
-#define cpu_relax()		rep_nop()
+static __always_inline void cpu_relax(void)
+{
+	if (time_travel_mode == TT_MODE_INFCPU ||
+	    time_travel_mode == TT_MODE_EXTERNAL)
+		time_travel_ndelay(1);
+	else
+		rep_nop();
+}
 
 #define task_pt_regs(t) (&(t)->thread.regs)