lguest: Tell Guest net not to notify us on every packet xmit
virtio_ring has the ability to suppress notifications. This prevents
a guest exit for every packet, but we need to set a timer on packet
receipt to re-check if there were any remaining packets.
Here are the times for 1G TCP Guest->Host with different timeout
settings (it matters because the TCP window doesn't grow big enough to
fill the entire buffer):
Timeout value Seconds Xmit/Recv/Timeout
None (before) 25.3784 xmit 7750233 recv 1
2500 usec 62.5119 xmit 207020 recv 2 timeout 207020
1000 usec 34.5379 xmit 207003 recv 2 timeout 207003
750 usec 29.2305 xmit 207002 recv 1 timeout 207002
500 usec 19.1887 xmit 561141 recv 1 timeout 559657
250 usec 20.0465 xmit 214128 recv 2 timeout 214110
100 usec 19.2583 xmit 561621 recv 1 timeout 560153
(Note that these values are sensitive to the GSO patches which come
later, and probably other traffic-related variables, so take with a
large grain of salt).
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
diff --git a/Documentation/lguest/lguest.c b/Documentation/lguest/lguest.c
index 46f4c5b..018472c 100644
--- a/Documentation/lguest/lguest.c
+++ b/Documentation/lguest/lguest.c
@@ -36,6 +36,7 @@
#include <sched.h>
#include <limits.h>
#include <stddef.h>
+#include <signal.h>
#include "linux/lguest_launcher.h"
#include "linux/virtio_config.h"
#include "linux/virtio_net.h"
@@ -81,6 +82,8 @@
static void *guest_base;
/* The maximum guest physical address allowed, and maximum possible. */
static unsigned long guest_limit, guest_max;
+/* The pipe for signal hander to write to. */
+static int timeoutpipe[2];
/* a per-cpu variable indicating whose vcpu is currently running */
static unsigned int __thread cpu_id;
@@ -156,11 +159,14 @@
/* Last available index we saw. */
u16 last_avail_idx;
- /* The routine to call when the Guest pings us. */
- void (*handle_output)(int fd, struct virtqueue *me);
+ /* The routine to call when the Guest pings us, or timeout. */
+ void (*handle_output)(int fd, struct virtqueue *me, bool timeout);
/* Outstanding buffers */
unsigned int inflight;
+
+ /* Is this blocked awaiting a timer? */
+ bool blocked;
};
/* Remember the arguments to the program so we can "reboot" */
@@ -874,7 +880,7 @@
/* Handling output for console is simple: we just get all the output buffers
* and write them to stdout. */
-static void handle_console_output(int fd, struct virtqueue *vq)
+static void handle_console_output(int fd, struct virtqueue *vq, bool timeout)
{
unsigned int head, out, in;
int len;
@@ -889,6 +895,21 @@
}
}
+static void block_vq(struct virtqueue *vq)
+{
+ struct itimerval itm;
+
+ vq->vring.used->flags |= VRING_USED_F_NO_NOTIFY;
+ vq->blocked = true;
+
+ itm.it_interval.tv_sec = 0;
+ itm.it_interval.tv_usec = 0;
+ itm.it_value.tv_sec = 0;
+ itm.it_value.tv_usec = 500;
+
+ setitimer(ITIMER_REAL, &itm, NULL);
+}
+
/*
* The Network
*
@@ -896,9 +917,9 @@
* and write them (ignoring the first element) to this device's file descriptor
* (/dev/net/tun).
*/
-static void handle_net_output(int fd, struct virtqueue *vq)
+static void handle_net_output(int fd, struct virtqueue *vq, bool timeout)
{
- unsigned int head, out, in;
+ unsigned int head, out, in, num = 0;
int len;
struct iovec iov[vq->vring.num];
@@ -912,7 +933,12 @@
(void)convert(&iov[0], struct virtio_net_hdr);
len = writev(vq->dev->fd, iov+1, out-1);
add_used_and_trigger(fd, vq, head, len);
+ num++;
}
+
+ /* Block further kicks and set up a timer if we saw anything. */
+ if (!timeout && num)
+ block_vq(vq);
}
/* This is where we handle a packet coming in from the tun device to our
@@ -967,18 +993,18 @@
/*L:215 This is the callback attached to the network and console input
* virtqueues: it ensures we try again, in case we stopped console or net
* delivery because Guest didn't have any buffers. */
-static void enable_fd(int fd, struct virtqueue *vq)
+static void enable_fd(int fd, struct virtqueue *vq, bool timeout)
{
add_device_fd(vq->dev->fd);
/* Tell waker to listen to it again */
write(waker_fd, &vq->dev->fd, sizeof(vq->dev->fd));
}
-static void net_enable_fd(int fd, struct virtqueue *vq)
+static void net_enable_fd(int fd, struct virtqueue *vq, bool timeout)
{
/* We don't need to know again when Guest refills receive buffer. */
vq->vring.used->flags |= VRING_USED_F_NO_NOTIFY;
- enable_fd(fd, vq);
+ enable_fd(fd, vq, timeout);
}
/* When the Guest tells us they updated the status field, we handle it. */
@@ -1047,7 +1073,7 @@
if (strcmp(vq->dev->name, "console") != 0)
verbose("Output to %s\n", vq->dev->name);
if (vq->handle_output)
- vq->handle_output(fd, vq);
+ vq->handle_output(fd, vq, false);
return;
}
}
@@ -1061,6 +1087,29 @@
strnlen(from_guest_phys(addr), guest_limit - addr));
}
+static void handle_timeout(int fd)
+{
+ char buf[32];
+ struct device *i;
+ struct virtqueue *vq;
+
+ /* Clear the pipe */
+ read(timeoutpipe[0], buf, sizeof(buf));
+
+ /* Check each device and virtqueue: flush blocked ones. */
+ for (i = devices.dev; i; i = i->next) {
+ for (vq = i->vq; vq; vq = vq->next) {
+ if (!vq->blocked)
+ continue;
+
+ vq->vring.used->flags &= ~VRING_USED_F_NO_NOTIFY;
+ vq->blocked = false;
+ if (vq->handle_output)
+ vq->handle_output(fd, vq, true);
+ }
+ }
+}
+
/* This is called when the Waker wakes us up: check for incoming file
* descriptors. */
static void handle_input(int fd)
@@ -1071,9 +1120,14 @@
for (;;) {
struct device *i;
fd_set fds = devices.infds;
+ int num;
+ num = select(devices.max_infd+1, &fds, NULL, NULL, &poll);
+ /* Could get interrupted */
+ if (num < 0)
+ continue;
/* If nothing is ready, we're done. */
- if (select(devices.max_infd+1, &fds, NULL, NULL, &poll) == 0)
+ if (num == 0)
break;
/* Otherwise, call the device(s) which have readable file
@@ -1097,6 +1151,10 @@
write(waker_fd, &dev_fd, sizeof(dev_fd));
}
}
+
+ /* Is this the timeout fd? */
+ if (FD_ISSET(timeoutpipe[0], &fds))
+ handle_timeout(fd);
}
}
@@ -1145,7 +1203,7 @@
/* Each device descriptor is followed by the description of its virtqueues. We
* specify how many descriptors the virtqueue is to have. */
static void add_virtqueue(struct device *dev, unsigned int num_descs,
- void (*handle_output)(int fd, struct virtqueue *me))
+ void (*handle_output)(int, struct virtqueue *, bool))
{
unsigned int pages;
struct virtqueue **i, *vq = malloc(sizeof(*vq));
@@ -1161,6 +1219,7 @@
vq->last_avail_idx = 0;
vq->dev = dev;
vq->inflight = 0;
+ vq->blocked = false;
/* Initialize the configuration. */
vq->config.num = num_descs;
@@ -1293,6 +1352,24 @@
}
/*:*/
+static void timeout_alarm(int sig)
+{
+ write(timeoutpipe[1], "", 1);
+}
+
+static void setup_timeout(void)
+{
+ if (pipe(timeoutpipe) != 0)
+ err(1, "Creating timeout pipe");
+
+ if (fcntl(timeoutpipe[1], F_SETFL,
+ fcntl(timeoutpipe[1], F_GETFL) | O_NONBLOCK) != 0)
+ err(1, "Making timeout pipe nonblocking");
+
+ add_device_fd(timeoutpipe[0]);
+ signal(SIGALRM, timeout_alarm);
+}
+
/*M:010 Inter-guest networking is an interesting area. Simplest is to have a
* --sharenet=<name> option which opens or creates a named pipe. This can be
* used to send packets to another guest in a 1:1 manner.
@@ -1653,7 +1730,7 @@
}
/* When the Guest submits some I/O, we just need to wake the I/O thread. */
-static void handle_virtblk_output(int fd, struct virtqueue *vq)
+static void handle_virtblk_output(int fd, struct virtqueue *vq, bool timeout)
{
struct vblk_info *vblk = vq->dev->priv;
char c = 0;
@@ -1824,7 +1901,7 @@
/* ERESTART means that we need to reboot the guest */
} else if (errno == ERESTART) {
restart_guest();
- /* EAGAIN means the Waker wanted us to look at some input.
+ /* EAGAIN means a signal (timeout).
* Anything else means a bug or incompatible change. */
} else if (errno != EAGAIN)
err(1, "Running guest failed");
@@ -1948,6 +2025,9 @@
/* We always have a console device */
setup_console();
+ /* We can timeout waiting for Guest network transmit. */
+ setup_timeout();
+
/* Now we load the kernel */
start = load_kernel(open_or_die(argv[optind+1], O_RDONLY));