This patch adds the virtio balloon driver. It doesn't actually reduce the RSS
size when using KVM (although it does when using -no-kvm). This is because
even with MMU notifiers, we're holding a reference to each page in the shadow
page table. Once we switch to not taking a reference for pages in the shadow
page table, the balloon driver should work as expected.
Signed-off-by: Anthony Liguori <[EMAIL PROTECTED]>
diff --git a/qemu/Makefile.target b/qemu/Makefile.target
index a86464f..cd985f6 100644
--- a/qemu/Makefile.target
+++ b/qemu/Makefile.target
@@ -598,7 +598,7 @@ OBJS += rtl8139.o
OBJS += e1000.o
# virtio devices
-OBJS += virtio.o virtio-net.o virtio-blk.o
+OBJS += virtio.o virtio-net.o virtio-blk.o virtio-balloon.o
OBJS += device-hotplug.o
diff --git a/qemu/hw/pc.c b/qemu/hw/pc.c
index 3a8269b..b19e558 100644
--- a/qemu/hw/pc.c
+++ b/qemu/hw/pc.c
@@ -1161,6 +1161,9 @@ static void pc_init1(ram_addr_t ram_size, int
vga_ram_size,
extboot_init(info->bdrv, 1);
}
+
+ if (pci_enabled)
+ virtio_balloon_init(pci_bus);
}
static void pc_init_pci(ram_addr_t ram_size, int vga_ram_size,
diff --git a/qemu/hw/pc.h b/qemu/hw/pc.h
index c284bf1..5b68d69 100644
--- a/qemu/hw/pc.h
+++ b/qemu/hw/pc.h
@@ -164,4 +164,7 @@ void *virtio_blk_init(PCIBus *bus, uint16_t vendor,
uint16_t device,
void extboot_init(BlockDriverState *bs, int cmd);
+/* virtio-balloon.h */
+void *virtio_balloon_init(PCIBus *bus);
+
#endif
diff --git a/qemu/hw/virtio-balloon.c b/qemu/hw/virtio-balloon.c
new file mode 100644
index 0000000..457b88f
--- /dev/null
+++ b/qemu/hw/virtio-balloon.c
@@ -0,0 +1,194 @@
+/*
+ * Virtio Block Device
+ *
+ * Copyright IBM, Corp. 2008
+ *
+ * Authors:
+ * Anthony Liguori <[EMAIL PROTECTED]>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See
+ * the COPYING file in the top-level directory.
+ *
+ */
+
+#include "qemu-common.h"
+#include "virtio.h"
+#include "pc.h"
+#include "sysemu.h"
+#include "cpu.h"
+#include "balloon.h"
+#include "virtio-balloon.h"
+#include "qemu-kvm.h"
+
+#if defined(__linux__)
+#include <sys/mman.h>
+#endif
+
+typedef struct VirtIOBalloon
+{
+ VirtIODevice vdev;
+ VirtQueue *ivq, *dvq;
+ uint32_t num_pages;
+ uint32_t actual;
+} VirtIOBalloon;
+
+static VirtIOBalloon *to_virtio_balloon(VirtIODevice *vdev)
+{
+ return (VirtIOBalloon *)vdev;
+}
+
+static void balloon_page(void *addr, int deflate)
+{
+#if defined(__linux__)
+ if (!kvm_enabled() || !qemu_kvm_has_mmu_notifiers())
+ madvise(addr, TARGET_PAGE_SIZE,
+ deflate ? MADV_WILLNEED : MADV_DONTNEED);
+#endif
+}
+
+/* FIXME: once we do a virtio refactoring, this will get subsumed into common
+ * code */
+static size_t memcpy_from_iovector(void *data, size_t offset, size_t size,
+ struct iovec *iov, int iovlen)
+{
+ int i;
+ uint8_t *ptr = data;
+ size_t iov_off = 0;
+ size_t data_off = 0;
+
+ for (i = 0; i < iovlen && size; i++) {
+ if (offset < (iov_off + iov[i].iov_len)) {
+ size_t len = MIN((iov_off + iov[i].iov_len) - offset , size);
+
+ memcpy(ptr + data_off, iov[i].iov_base + (offset - iov_off), len);
+
+ data_off += len;
+ offset += len;
+ size -= len;
+ }
+
+ iov_off += iov[i].iov_len;
+ }
+
+ return data_off;
+}
+
+static void virtio_balloon_handle_output(VirtIODevice *vdev, VirtQueue *vq)
+{
+ VirtIOBalloon *s = to_virtio_balloon(vdev);
+ VirtQueueElement elem;
+
+ while (virtqueue_pop(vq, &elem)) {
+ size_t offset = 0;
+ uint32_t pfn;
+
+ while (memcpy_from_iovector(&pfn, offset, 4,
+ elem.out_sg, elem.out_num) == 4) {
+ ram_addr_t pa;
+ ram_addr_t addr;
+
+ pa = (ram_addr_t)ldl_p(&pfn) << TARGET_PAGE_BITS;
+ offset += 4;
+
+ addr = cpu_get_physical_page_desc(pa);
+ if ((addr & ~TARGET_PAGE_MASK) != IO_MEM_RAM)
+ continue;
+
+ balloon_page(phys_ram_base + addr, !!(vq == s->dvq));
+ }
+
+ virtqueue_push(vq, &elem, offset);
+ virtio_notify(vdev, vq);
+ }
+}
+
+static void virtio_balloon_get_config(VirtIODevice *vdev, uint8_t *config_data)
+{
+ VirtIOBalloon *dev = to_virtio_balloon(vdev);
+ struct virtio_balloon_config config;
+
+ config.num_pages = cpu_to_le32(dev->num_pages);
+ config.actual = cpu_to_le32(dev->actual);
+
+ memcpy(config_data, &config, 8);
+}
+
+static void virtio_balloon_set_config(VirtIODevice *vdev,
+ const uint8_t *config_data)
+{
+ VirtIOBalloon *dev = to_virtio_balloon(vdev);
+ struct virtio_balloon_config config;
+ memcpy(&config, config_data, 8);
+ dev->actual = config.actual;
+}
+
+static uint32_t virtio_balloon_get_features(VirtIODevice *vdev)
+{
+ return 0;
+}
+
+static ram_addr_t virtio_balloon_to_target(void *opaque, ram_addr_t target)
+{
+ VirtIOBalloon *dev = opaque;
+
+ if (target > ram_size)
+ target = ram_size;
+
+ if (target) {
+ dev->num_pages = (ram_size - target) >> TARGET_PAGE_BITS;
+ virtio_notify_config(&dev->vdev);
+ }
+
+ return ram_size - (dev->actual << TARGET_PAGE_BITS);
+}
+
+static void virtio_balloon_save(QEMUFile *f, void *opaque)
+{
+ VirtIOBalloon *s = opaque;
+
+ virtio_save(&s->vdev, f);
+
+ qemu_put_be32(f, s->num_pages);
+ qemu_put_be32(f, s->actual);
+}
+
+static int virtio_balloon_load(QEMUFile *f, void *opaque, int version_id)
+{
+ VirtIOBalloon *s = opaque;
+
+ if (version_id != 1)
+ return -EINVAL;
+
+ virtio_load(&s->vdev, f);
+
+ s->num_pages = qemu_get_be32(f);
+ s->actual = qemu_get_be32(f);
+
+ return 0;
+}
+
+void *virtio_balloon_init(PCIBus *bus)
+{
+ VirtIOBalloon *s;
+
+ s = (VirtIOBalloon *)virtio_init_pci(bus, "virtio-balloon",
+ 6900, 0x1002,
+ 0, VIRTIO_ID_BALLOON,
+ 0x05, 0x00, 0x00,
+ 8, sizeof(VirtIOBalloon));
+ if (s == NULL)
+ return NULL;
+
+ s->vdev.get_config = virtio_balloon_get_config;
+ s->vdev.set_config = virtio_balloon_set_config;
+ s->vdev.get_features = virtio_balloon_get_features;
+
+ s->ivq = virtio_add_queue(&s->vdev, 128, virtio_balloon_handle_output);
+ s->dvq = virtio_add_queue(&s->vdev, 128, virtio_balloon_handle_output);
+
+ qemu_add_balloon_handler(virtio_balloon_to_target, s);
+
+ register_savevm("virtio-balloon", -1, 1, virtio_balloon_save,
virtio_balloon_load, s);
+
+ return &s->vdev;
+}
diff --git a/qemu/hw/virtio-balloon.h b/qemu/hw/virtio-balloon.h
new file mode 100644
index 0000000..27d6985
--- /dev/null
+++ b/qemu/hw/virtio-balloon.h
@@ -0,0 +1,34 @@
+/*
+ * Virtio Support
+ *
+ * Copyright IBM, Corp. 2007-2008
+ *
+ * Authors:
+ * Anthony Liguori <[EMAIL PROTECTED]>
+ * Rusty Russell <[EMAIL PROTECTED]>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2. See
+ * the COPYING file in the top-level directory.
+ *
+ */
+
+#ifndef _QEMU_VIRTIO_BALLOON_H
+#define _QEMU_VIRTIO_BALLOON_H
+
+/* from Linux's linux/virtio_balloon.h */
+
+/* The ID for virtio_balloon */
+#define VIRTIO_ID_BALLOON 5
+
+/* The feature bitmap for virtio balloon */
+#define VIRTIO_BALLOON_F_MUST_TELL_HOST 0 /* Tell before reclaiming
pages */
+
+struct virtio_balloon_config
+{
+ /* Number of pages host wants Guest to give up. */
+ uint32_t num_pages;
+ /* Number of pages we've actually got in balloon. */
+ uint32_t actual;
+};
+
+#endif
diff --git a/qemu/monitor.c b/qemu/monitor.c
index 2240d8e..b8c4fde 100644
--- a/qemu/monitor.c
+++ b/qemu/monitor.c
@@ -1411,7 +1411,9 @@ static void do_info_balloon(void)
ram_addr_t actual;
actual = qemu_balloon_status();
- if (actual == 0)
+ if (kvm_enabled() && !qemu_kvm_has_mmu_notifiers())
+ term_printf("Using KVM without MMU-notifiers, ballooning disabled\n");
+ else if (actual == 0)
term_printf("Ballooning not activated in VM\n");
else
term_printf("balloon: actual=%d\n", (int)(actual >> 20));
diff --git a/qemu/qemu-kvm.h b/qemu/qemu-kvm.h
index 7e28428..088bae3 100644
--- a/qemu/qemu-kvm.h
+++ b/qemu/qemu-kvm.h
@@ -110,10 +110,12 @@ extern kvm_context_t kvm_context;
#define kvm_enabled() (kvm_allowed)
#define qemu_kvm_irqchip_in_kernel() kvm_irqchip_in_kernel(kvm_context)
#define qemu_kvm_pit_in_kernel() kvm_pit_in_kernel(kvm_context)
+#define qemu_kvm_has_mmu_notifiers() kvm_has_mmu_notifiers(kvm_context)
#else
#define kvm_enabled() (0)
#define qemu_kvm_irqchip_in_kernel() (0)
#define qemu_kvm_pit_in_kernel() (0)
+#define qemu_kvm_has_mmu_notifiers() (0)
#endif
void kvm_mutex_unlock(void);
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
