v3vee-$(V3_CONFIG_EXT_SELECTIVE_SYSCALL_EXIT) += iface-syscall.o
+v3vee-$(V3_CONFIG_HOST_PCI) += iface-host-pci.o
+
v3vee-objs := $(v3vee-y) ../libv3vee.a
obj-m := v3vee.o
struct palacios_console * cons = (struct palacios_console *) console;
cons->open = 0;
+
+ remove_guest_ctrl(cons->guest, V3_VM_CONSOLE_CONNECT);
+ deinit_queue(cons->queue);
+
+ kfree(cons);
}
--- /dev/null
+/* Linux host side PCI passthrough support
+ * Jack Lange <jacklange@cs.pitt.edu>, 2012
+ */
+
+#include <linux/pci.h>
+#include <linux/iommu.h>
+#include <linux/interrupt.h>
+#include <linux/version.h>
+
+
+#define PCI_HDR_SIZE 256
+
+
+static int setup_hw_pci_dev(struct host_pci_device * host_dev) {
+ int ret = 0;
+ struct pci_dev * dev = NULL;
+ struct v3_host_pci_dev * v3_dev = &(host_dev->v3_dev);
+
+ dev = pci_get_bus_and_slot(host_dev->hw_dev.bus,
+ host_dev->hw_dev.devfn);
+
+
+ if (dev == NULL) {
+ printk("Could not find HW pci device (bus=%d, devfn=%d)\n",
+ host_dev->hw_dev.bus, host_dev->hw_dev.devfn);
+ return -1;
+ }
+
+ // record pointer in dev state
+ host_dev->hw_dev.dev = dev;
+
+ host_dev->hw_dev.intx_disabled = 1;
+ spin_lock_init(&(host_dev->hw_dev.intx_lock));
+
+ if (pci_enable_device(dev)) {
+ printk("Could not enable Device\n");
+ return -1;
+ }
+
+ ret = pci_request_regions(dev, "v3vee");
+ if (ret != 0) {
+ printk("Could not reservce PCI regions\n");
+ return -1;
+ }
+
+
+ pci_reset_function(host_dev->hw_dev.dev);
+ pci_save_state(host_dev->hw_dev.dev);
+
+
+ {
+ int i = 0;
+ for (i = 0; i < DEVICE_COUNT_RESOURCE; i++) {
+ printk("Resource %d\n", i);
+ printk("\tflags = 0x%lx\n", pci_resource_flags(dev, i));
+ printk("\t name=%s, start=%lx, size=%d\n",
+ host_dev->hw_dev.dev->resource[i].name, (uintptr_t)pci_resource_start(dev, i),
+ (u32)pci_resource_len(dev, i));
+
+ }
+
+ printk("Rom BAR=%d\n", dev->rom_base_reg);
+ }
+
+ /* Cache first 6 BAR regs */
+ {
+ int i = 0;
+
+ for (i = 0; i < 6; i++) {
+ struct v3_host_pci_bar * bar = &(v3_dev->bars[i]);
+ unsigned long flags;
+
+ bar->size = pci_resource_len(dev, i);
+ bar->addr = pci_resource_start(dev, i);
+ flags = pci_resource_flags(dev, i);
+
+ if (flags & IORESOURCE_IO) {
+ bar->type = PT_BAR_IO;
+ } else if (flags & IORESOURCE_MEM) {
+ if (flags & IORESOURCE_MEM_64) {
+ struct v3_host_pci_bar * hi_bar = &(v3_dev->bars[i + 1]);
+
+ bar->type = PT_BAR_MEM64_LO;
+
+ hi_bar->type = PT_BAR_MEM64_HI;
+ hi_bar->size = bar->size;
+ hi_bar->addr = bar->addr;
+ hi_bar->cacheable = ((flags & IORESOURCE_CACHEABLE) != 0);
+ hi_bar->prefetchable = ((flags & IORESOURCE_PREFETCH) != 0);
+
+ i++;
+ } else if (flags & IORESOURCE_DMA) {
+ bar->type = PT_BAR_MEM24;
+ } else {
+ bar->type = PT_BAR_MEM32;
+ }
+
+ bar->cacheable = ((flags & IORESOURCE_CACHEABLE) != 0);
+ bar->prefetchable = ((flags & IORESOURCE_PREFETCH) != 0);
+
+ } else {
+ bar->type = PT_BAR_NONE;
+ }
+ }
+ }
+
+ /* Cache expansion rom bar */
+ {
+ struct resource * rom_res = &(dev->resource[PCI_ROM_RESOURCE]);
+ int rom_size = pci_resource_len(dev, PCI_ROM_RESOURCE);
+
+ if (rom_size > 0) {
+ unsigned long flags;
+
+ v3_dev->exp_rom.size = rom_size;
+ v3_dev->exp_rom.addr = pci_resource_start(dev, PCI_ROM_RESOURCE);
+ flags = pci_resource_flags(dev, PCI_ROM_RESOURCE);
+
+ v3_dev->exp_rom.type = PT_EXP_ROM;
+
+ v3_dev->exp_rom.exp_rom_enabled = rom_res->flags & IORESOURCE_ROM_ENABLE;
+ }
+ }
+
+ /* Cache entire configuration space */
+ {
+ int m = 0;
+
+ // Copy the configuration space to the local cached version
+ for (m = 0; m < PCI_HDR_SIZE; m += 4) {
+ pci_read_config_dword(dev, m, (u32 *)&(v3_dev->cfg_space[m]));
+ }
+ }
+
+
+ /* HARDCODED for now but this will need to depend on IOMMU support detection */
+ if (iommu_found()) {
+ printk("Setting host PCI device (%s) as IOMMU\n", host_dev->name);
+ v3_dev->iface = IOMMU;
+ } else {
+ printk("Setting host PCI device (%s) as SYMBIOTIC\n", host_dev->name);
+ v3_dev->iface = SYMBIOTIC;
+ }
+
+ return 0;
+
+}
+
+
+
+static irqreturn_t host_pci_intx_irq_handler(int irq, void * priv_data) {
+ struct host_pci_device * host_dev = priv_data;
+
+ // printk("Host PCI IRQ handler (%d)\n", irq);
+
+ spin_lock(&(host_dev->hw_dev.intx_lock));
+ disable_irq_nosync(irq);
+ host_dev->hw_dev.intx_disabled = 1;
+ spin_unlock(&(host_dev->hw_dev.intx_lock));
+
+ V3_host_pci_raise_irq(&(host_dev->v3_dev), 0);
+
+ return IRQ_HANDLED;
+}
+
+
+
+static irqreturn_t host_pci_msi_irq_handler(int irq, void * priv_data) {
+ struct host_pci_device * host_dev = priv_data;
+ // printk("Host PCI MSI IRQ Handler (%d)\n", irq);
+
+ V3_host_pci_raise_irq(&(host_dev->v3_dev), 0);
+
+ return IRQ_HANDLED;
+}
+
+static irqreturn_t host_pci_msix_irq_handler(int irq, void * priv_data) {
+ struct host_pci_device * host_dev = priv_data;
+ int i = 0;
+
+ // printk("Host PCI MSIX IRQ Handler (%d)\n", irq);
+
+ // find vector index
+ for (i = 0; i < host_dev->hw_dev.num_msix_vecs; i++) {
+ if (irq == host_dev->hw_dev.msix_entries[i].vector) {
+ V3_host_pci_raise_irq(&(host_dev->v3_dev), i);
+ } else {
+ printk("Error Could not find matching MSIX vector for IRQ %d\n", irq);
+ }
+ }
+ return IRQ_HANDLED;
+}
+
+
+static int hw_pci_cmd(struct host_pci_device * host_dev, host_pci_cmd_t cmd, u64 arg) {
+ //struct v3_host_pci_dev * v3_dev = &(host_dev->v3_dev);
+ struct pci_dev * dev = host_dev->hw_dev.dev;
+
+ switch (cmd) {
+ case HOST_PCI_CMD_DMA_DISABLE:
+ printk("Passthrough PCI device disabling BMDMA\n");
+ pci_clear_master(host_dev->hw_dev.dev);
+ break;
+ case HOST_PCI_CMD_DMA_ENABLE:
+ printk("Passthrough PCI device Enabling BMDMA\n");
+ pci_set_master(host_dev->hw_dev.dev);
+ break;
+
+ case HOST_PCI_CMD_INTX_DISABLE:
+ printk("Passthrough PCI device disabling INTx IRQ\n");
+
+ disable_irq(dev->irq);
+ free_irq(dev->irq, (void *)host_dev);
+
+ break;
+ case HOST_PCI_CMD_INTX_ENABLE:
+ printk("Passthrough PCI device Enabling INTx IRQ\n");
+
+ if (request_threaded_irq(dev->irq, NULL, host_pci_intx_irq_handler,
+ IRQF_ONESHOT, "V3Vee_Host_PCI_INTx", (void *)host_dev)) {
+ printk("ERROR Could not assign IRQ to host PCI device (%s)\n", host_dev->name);
+ }
+
+ break;
+
+ case HOST_PCI_CMD_MSI_DISABLE:
+ printk("Passthrough PCI device Disabling MSIs\n");
+
+ disable_irq(dev->irq);
+ free_irq(dev->irq, (void *)host_dev);
+
+ pci_disable_msi(dev);
+
+ break;
+ case HOST_PCI_CMD_MSI_ENABLE:
+ printk("Passthrough PCI device Enabling MSI\n");
+
+ if (!dev->msi_enabled) {
+ pci_enable_msi(dev);
+
+ if (request_irq(dev->irq, host_pci_msi_irq_handler,
+ 0, "V3Vee_host_PCI_MSI", (void *)host_dev)) {
+ printk("Error Requesting IRQ %d for Passthrough MSI IRQ\n", dev->irq);
+ }
+ }
+
+ break;
+
+
+
+ case HOST_PCI_CMD_MSIX_ENABLE: {
+ int i = 0;
+
+ printk("Passthrough PCI device Enabling MSIX\n");
+ host_dev->hw_dev.num_msix_vecs = arg;;
+ host_dev->hw_dev.msix_entries = kcalloc(host_dev->hw_dev.num_msix_vecs,
+ sizeof(struct msix_entry), GFP_KERNEL);
+
+ for (i = 0; i < host_dev->hw_dev.num_msix_vecs; i++) {
+ host_dev->hw_dev.msix_entries[i].entry = i;
+ }
+
+ pci_enable_msix(dev, host_dev->hw_dev.msix_entries,
+ host_dev->hw_dev.num_msix_vecs);
+
+ for (i = 0; i < host_dev->hw_dev.num_msix_vecs; i++) {
+ if (request_irq(host_dev->hw_dev.msix_entries[i].vector,
+ host_pci_msix_irq_handler,
+ 0, "V3VEE_host_PCI_MSIX", (void *)host_dev)) {
+ printk("Error requesting IRQ %d for Passthrough MSIX IRQ\n",
+ host_dev->hw_dev.msix_entries[i].vector);
+ }
+ }
+
+ break;
+ }
+
+ case HOST_PCI_CMD_MSIX_DISABLE: {
+ int i = 0;
+
+ printk("Passthrough PCI device Disabling MSIX\n");
+
+ for (i = 0; i < host_dev->hw_dev.num_msix_vecs; i++) {
+ disable_irq(host_dev->hw_dev.msix_entries[i].vector);
+ }
+
+ for (i = 0; i < host_dev->hw_dev.num_msix_vecs; i++) {
+ free_irq(host_dev->hw_dev.msix_entries[i].vector, (void *)host_dev);
+ }
+
+ host_dev->hw_dev.num_msix_vecs = 0;
+ kfree(host_dev->hw_dev.msix_entries);
+
+ pci_disable_msix(dev);
+
+ break;
+ }
+ default:
+ printk("Error: unhandled passthrough PCI command: %d\n", cmd);
+ return -1;
+
+ }
+
+ return 0;
+}
+
+
+static int hw_ack_irq(struct host_pci_device * host_dev, u32 vector) {
+ struct pci_dev * dev = host_dev->hw_dev.dev;
+ unsigned long flags;
+
+ // printk("Acking IRQ vector %d\n", vector);
+
+ spin_lock_irqsave(&(host_dev->hw_dev.intx_lock), flags);
+ // printk("Enabling IRQ %d\n", dev->irq);
+ enable_irq(dev->irq);
+ host_dev->hw_dev.intx_disabled = 0;
+ spin_unlock_irqrestore(&(host_dev->hw_dev.intx_lock), flags);
+
+ return 0;
+}
+
+
+
+
+static int reserve_hw_pci_dev(struct host_pci_device * host_dev, void * v3_ctx) {
+ int ret = 0;
+ unsigned long flags;
+ struct v3_host_pci_dev * v3_dev = &(host_dev->v3_dev);
+ struct pci_dev * dev = host_dev->hw_dev.dev;
+
+ spin_lock_irqsave(&lock, flags);
+ if (host_dev->hw_dev.in_use == 0) {
+ host_dev->hw_dev.in_use = 1;
+ } else {
+ ret = -1;
+ }
+ spin_unlock_irqrestore(&lock, flags);
+
+
+ if (v3_dev->iface == IOMMU) {
+ struct v3_guest_mem_region region;
+ int flags = 0;
+
+ host_dev->hw_dev.iommu_domain = iommu_domain_alloc();
+
+ if (V3_get_guest_mem_region(v3_ctx, ®ion) == -1) {
+ printk("Error getting VM memory region for IOMMU support\n");
+ return -1;
+ }
+
+ printk("Memory region: start=%p, end=%p\n", (void *)region.start, (void *)region.end);
+
+
+ flags = IOMMU_READ | IOMMU_WRITE; // Need to see what IOMMU_CACHE means
+
+ /* This version could be wrong */
+#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,38)
+ // Guest VAs start at zero and go to end of memory
+ iommu_map_range(host_dev->hw_dev.iommu_domain, 0, region.start, (region.end - region.start), flags);
+#else
+ /* Linux actually made the interface worse... Now you can only map memory in powers of 2 (meant to only be pages...) */
+ {
+ u64 size = region.end - region.start;
+ u32 page_size = 512 * 4096; // assume large 64bit pages (2MB)
+ u64 dpa = 0; // same as gpa
+ u64 hpa = region.start;
+
+ do {
+ if (size < page_size) {
+ page_size = 4096; // less than a 2MB granularity, so we switch to small pages (4KB)
+ }
+
+ printk("Mapping IOMMU region dpa=%p hpa=%p (size=%d)\n", (void *)dpa, (void *)hpa, page_size);
+
+ if (iommu_map(host_dev->hw_dev.iommu_domain, dpa, hpa,
+ get_order(page_size), flags)) {
+ printk("ERROR: Could not map sub region (DPA=%p) (HPA=%p) (order=%d)\n",
+ (void *)dpa, (void *)hpa, get_order(page_size));
+ break;
+ }
+
+ hpa += page_size;
+ dpa += page_size;
+
+ size -= page_size;
+ } while (size);
+ }
+#endif
+
+ if (iommu_attach_device(host_dev->hw_dev.iommu_domain, &(dev->dev))) {
+ printk("ERROR attaching host PCI device to IOMMU domain\n");
+ }
+
+ }
+
+
+ printk("Requesting Threaded IRQ handler for IRQ %d\n", dev->irq);
+ // setup regular IRQs until advanced IRQ mechanisms are enabled
+ if (request_threaded_irq(dev->irq, NULL, host_pci_intx_irq_handler,
+ IRQF_ONESHOT, "V3Vee_Host_PCI_INTx", (void *)host_dev)) {
+ printk("ERROR Could not assign IRQ to host PCI device (%s)\n", host_dev->name);
+ }
+
+
+
+
+ return ret;
+}
+
+
+
+static int write_hw_pci_config(struct host_pci_device * host_dev, u32 reg, void * data, u32 length) {
+ struct pci_dev * dev = host_dev->hw_dev.dev;
+
+ if (reg < 64) {
+ return 0;
+ }
+
+ if (length == 1) {
+ pci_write_config_byte(dev, reg, *(u8 *)data);
+ } else if (length == 2) {
+ pci_write_config_word(dev, reg, *(u16 *)data);
+ } else if (length == 4) {
+ pci_write_config_dword(dev, reg, *(u32 *)data);
+ } else {
+ printk("Invalid length of host PCI config update\n");
+ return -1;
+ }
+
+ return 0;
+}
+
+
+
+static int read_hw_pci_config(struct host_pci_device * host_dev, u32 reg, void * data, u32 length) {
+ struct pci_dev * dev = host_dev->hw_dev.dev;
+
+
+ if (length == 1) {
+ pci_read_config_byte(dev, reg, data);
+ } else if (length == 2) {
+ pci_read_config_word(dev, reg, data);
+ } else if (length == 4) {
+ pci_read_config_dword(dev, reg, data);
+ } else {
+ printk("Invalid length of host PCI config read\n");
+ return -1;
+ }
+
+
+ return 0;
+}
--- /dev/null
+/* Host PCI interface
+ * (c) Jack Lange, 2012
+ * jacklange@cs.pitt.edu
+ */
+
+#include <linux/uaccess.h>
+#include <linux/fs.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+
+#include "palacios.h"
+#include "linux-exts.h"
+
+
+#include <interfaces/host_pci.h>
+
+static struct list_head device_list;
+static spinlock_t lock;
+
+
+
+
+struct pci_dev;
+struct iommu_domain;
+
+struct host_pci_device {
+ char name[128];
+
+ enum {PASSTHROUGH, USER} type;
+
+ enum {INTX_IRQ, MSI_IRQ, MSIX_IRQ} irq_type;
+ uint32_t num_vecs;
+
+ union {
+ struct {
+ u8 in_use;
+ u8 iommu_enabled;
+
+ u32 bus;
+ u32 devfn;
+
+ spinlock_t intx_lock;
+ u8 intx_disabled;
+
+ u32 num_msix_vecs;
+ struct msix_entry * msix_entries;
+ struct iommu_domain * iommu_domain;
+
+ struct pci_dev * dev;
+ } hw_dev;
+
+ // struct user_dev_state user_dev;
+ };
+
+ struct v3_host_pci_dev v3_dev;
+
+ struct list_head dev_node;
+};
+
+
+//#include "iface-host-pci-user.h"
+#include "iface-host-pci-hw.h"
+
+
+static struct host_pci_device * find_dev_by_name(char * name) {
+ struct host_pci_device * dev = NULL;
+
+ list_for_each_entry(dev, &device_list, dev_node) {
+ if (strncmp(dev->name, name, 128) == 0) {
+ return dev;
+ }
+ }
+
+ return NULL;
+}
+
+
+
+static struct v3_host_pci_dev * request_pci_dev(char * url, void * v3_ctx) {
+
+ unsigned long flags;
+ struct host_pci_device * host_dev = NULL;
+
+ spin_lock_irqsave(&lock, flags);
+ host_dev = find_dev_by_name(url);
+ spin_unlock_irqrestore(&lock, flags);
+
+ if (host_dev == NULL) {
+ printk("Could not find host device (%s)\n", url);
+ return NULL;
+ }
+
+ if (host_dev->type == PASSTHROUGH) {
+ if (reserve_hw_pci_dev(host_dev, v3_ctx) == -1) {
+ printk("Could not reserve host device (%s)\n", url);
+ return NULL;
+ }
+ } else {
+ printk("Unsupported Host device type\n");
+ return NULL;
+ }
+
+
+
+ return &(host_dev->v3_dev);
+
+}
+
+
+static int host_pci_config_write(struct v3_host_pci_dev * v3_dev, unsigned int reg_num,
+ void * src, unsigned int length) {
+ struct host_pci_device * host_dev = v3_dev->host_data;
+
+ if (host_dev->type == PASSTHROUGH) {
+ return write_hw_pci_config(host_dev, reg_num, src, length);
+ }
+
+ printk("Error in config write handler\n");
+ return -1;
+}
+
+static int host_pci_config_read(struct v3_host_pci_dev * v3_dev, unsigned int reg_num,
+ void * dst, unsigned int length) {
+ struct host_pci_device * host_dev = v3_dev->host_data;
+
+ if (host_dev->type == PASSTHROUGH) {
+ return read_hw_pci_config(host_dev, reg_num, dst, length);
+ }
+
+ printk("Error in config read handler\n");
+ return -1;
+}
+
+
+static int host_pci_ack_irq(struct v3_host_pci_dev * v3_dev, unsigned int vector) {
+ struct host_pci_device * host_dev = v3_dev->host_data;
+
+ if (host_dev->type == PASSTHROUGH) {
+ return hw_ack_irq(host_dev, vector);
+ }
+
+ printk("Error in config irq ack handler\n");
+ return -1;
+}
+
+
+
+static int host_pci_cmd(struct v3_host_pci_dev * v3_dev, host_pci_cmd_t cmd, u64 arg) {
+ struct host_pci_device * host_dev = v3_dev->host_data;
+
+ if (host_dev->type == PASSTHROUGH) {
+ return hw_pci_cmd(host_dev, cmd, arg);
+ }
+
+ printk("Error in config pci cmd handler\n");
+ return -1;
+
+}
+
+static struct v3_host_pci_hooks pci_hooks = {
+ .request_device = request_pci_dev,
+ .config_write = host_pci_config_write,
+ .config_read = host_pci_config_read,
+ .ack_irq = host_pci_ack_irq,
+ .pci_cmd = host_pci_cmd,
+
+};
+
+
+
+static int register_pci_hw_dev(unsigned int cmd, unsigned long arg) {
+ void __user * argp = (void __user *)arg;
+ struct v3_hw_pci_dev hw_dev_arg ;
+ struct host_pci_device * host_dev = NULL;
+ unsigned long flags;
+ int ret = 0;
+
+ if (copy_from_user(&hw_dev_arg, argp, sizeof(struct v3_hw_pci_dev))) {
+ printk("%s(%d): copy from user error...\n", __FILE__, __LINE__);
+ return -EFAULT;
+ }
+
+ host_dev = kzalloc(sizeof(struct host_pci_device), GFP_KERNEL);
+
+
+ strncpy(host_dev->name, hw_dev_arg.name, 128);
+ host_dev->v3_dev.host_data = host_dev;
+
+
+ host_dev->type = PASSTHROUGH;
+ host_dev->hw_dev.bus = hw_dev_arg.bus;
+ host_dev->hw_dev.devfn = PCI_DEVFN(hw_dev_arg.dev, hw_dev_arg.func);
+
+
+ spin_lock_irqsave(&lock, flags);
+ if (!find_dev_by_name(hw_dev_arg.name)) {
+ list_add(&(host_dev->dev_node), &device_list);
+ ret = 1;
+ }
+ spin_unlock_irqrestore(&lock, flags);
+
+ if (ret == 0) {
+ // Error device already exists
+ kfree(host_dev);
+ return -EFAULT;
+ }
+
+
+ setup_hw_pci_dev(host_dev);
+
+ return 0;
+}
+
+
+static int register_pci_user_dev(unsigned int cmd, unsigned long arg) {
+ return 0;
+}
+
+
+
+
+static int host_pci_init( void ) {
+ INIT_LIST_HEAD(&(device_list));
+ spin_lock_init(&lock);
+
+ V3_Init_Host_PCI(&pci_hooks);
+
+
+ add_global_ctrl(V3_ADD_PCI_HW_DEV, register_pci_hw_dev);
+ add_global_ctrl(V3_ADD_PCI_USER_DEV, register_pci_user_dev);
+
+ return 0;
+}
+
+
+
+static struct linux_ext host_pci_ext = {
+ .name = "HOST_PCI",
+ .init = host_pci_init,
+};
+
+
+
+register_extension(&host_pci_ext);
-
#include "linux-exts.h"
/*
* This is a place holder to ensure that the _lnx_exts section gets created by gcc
*/
-
static struct {} null_ext __attribute__((__used__)) \
__attribute__((unused, __section__ ("_lnx_exts"), \
aligned(sizeof(void *))));
+
+
+/* */
+/* Global controls */
+/* */
+
+struct rb_root global_ctrls;
+
+static inline struct global_ctrl * __insert_global_ctrl(struct global_ctrl * ctrl) {
+ struct rb_node ** p = &(global_ctrls.rb_node);
+ struct rb_node * parent = NULL;
+ struct global_ctrl * tmp_ctrl = NULL;
+
+ while (*p) {
+ parent = *p;
+ tmp_ctrl = rb_entry(parent, struct global_ctrl, tree_node);
+
+ if (ctrl->cmd < tmp_ctrl->cmd) {
+ p = &(*p)->rb_left;
+ } else if (ctrl->cmd > tmp_ctrl->cmd) {
+ p = &(*p)->rb_right;
+ } else {
+ return tmp_ctrl;
+ }
+ }
+
+ rb_link_node(&(ctrl->tree_node), parent, p);
+
+ return NULL;
+}
+
+
+
+int add_global_ctrl(unsigned int cmd,
+ int (*handler)(unsigned int cmd, unsigned long arg)) {
+ struct global_ctrl * ctrl = kmalloc(sizeof(struct global_ctrl), GFP_KERNEL);
+
+ if (ctrl == NULL) {
+ printk("Error: Could not allocate global ctrl %d\n", cmd);
+ return -1;
+ }
+
+ ctrl->cmd = cmd;
+ ctrl->handler = handler;
+
+ if (__insert_global_ctrl(ctrl) != NULL) {
+ printk("Could not insert guest ctrl %d\n", cmd);
+ kfree(ctrl);
+ return -1;
+ }
+
+ rb_insert_color(&(ctrl->tree_node), &(global_ctrls));
+
+ return 0;
+}
+
+
+struct global_ctrl * get_global_ctrl(unsigned int cmd) {
+ struct rb_node * n = global_ctrls.rb_node;
+ struct global_ctrl * ctrl = NULL;
+
+ while (n) {
+ ctrl = rb_entry(n, struct global_ctrl, tree_node);
+
+ if (cmd < ctrl->cmd) {
+ n = n->rb_left;
+ } else if (cmd > ctrl->cmd) {
+ n = n->rb_right;
+ } else {
+ return ctrl;
+ }
+ }
+
+ return NULL;
+}
+
+
+
+
+
+/* */
+/* VM Controls */
+/* */
+
struct vm_ext {
struct linux_ext * impl;
void * vm_data;
return 0;
}
+
int init_lnx_extensions( void ) {
extern struct linux_ext * __start__lnx_exts[];
extern struct linux_ext * __stop__lnx_exts[];
while (tmp_ext != __stop__lnx_exts[0]) {
INFO("Cleaning up Linux Extension (%s)\n", tmp_ext->name);
+
if (tmp_ext->deinit != NULL) {
tmp_ext->deinit();
} else {
return 0;
}
+
#include "palacios.h"
-int add_mod_cmd(struct v3_guest * guest, unsigned int cmd,
- int (*handler)(struct v3_guest * guest,
- unsigned int cmd, unsigned long arg));
+
struct linux_ext {
+struct global_ctrl {
+ unsigned int cmd;
+
+ int (*handler)(unsigned int cmd, unsigned long arg);
+
+ struct rb_node tree_node;
+};
+
+int add_global_ctrl(unsigned int cmd,
+ int (*handler)(unsigned int cmd, unsigned long arg));
+
+struct global_ctrl * get_global_ctrl(unsigned int cmd);
+
+
+
#define register_extension(ext) \
static struct linux_ext * _lnx_ext \
__attribute__((used)) \
break;
}
- default:
- ERROR("\tUnhandled\n");
+
+ default: {
+ struct global_ctrl * ctrl = get_global_ctrl(ioctl);
+
+ if (ctrl) {
+ return ctrl->handler(ioctl, arg);
+ }
+
+ WARNING("\tUnhandled global ctrl cmd: %d\n", ioctl);
+
return -EINVAL;
+ }
}
return 0;
#define V3_FREE_GUEST 13
#define V3_ADD_MEMORY 50
+#define V3_ADD_PCI_HW_DEV 55
+#define V3_ADD_PCI_USER_DEV 56
/* VM Specific IOCTLs */
#define V3_VM_CONSOLE_CONNECT 20
} __attribute__((packed));
+struct v3_hw_pci_dev {
+ char name[128];
+ unsigned int bus;
+ unsigned int dev;
+ unsigned int func;
+} __attribute__((packed));
+
+struct v3_user_pci_dev {
+ char name[128];
+ unsigned short vendor_id;
+ unsigned short dev_id;
+} __attribute__((packed));
+
void * trace_malloc(size_t size, gfp_t flags);
spin_lock_init(&(queue->lock));
}
+void deinit_queue(struct gen_queue * queue) {
+ while (dequeue(queue)) {
+ ERROR("Freeing non-empty queue. PROBABLE MEMORY LEAK DETECTED\n");
+ }
+}
+
struct gen_queue * create_queue(unsigned int max_entries) {
struct gen_queue * tmp_queue = palacios_alloc(sizeof(struct gen_queue));
if (!tmp_queue) {
#define __PALACIOS_QUEUE_H__
+#include "palacios.h"
#include <linux/list.h>
#include <linux/spinlock.h>
+
struct queue_entry {
void * entry;
struct list_head node;
struct gen_queue * create_queue(unsigned int max_entries);
void init_queue(struct gen_queue * queue, unsigned int max_entries);
+void deinit_queue(struct gen_queue * queue);
+
int enqueue(struct gen_queue * queue, void * entry);
void * dequeue(struct gen_queue * queue);
}
+
+
static struct vm_ctrl * get_ctrl(struct v3_guest * guest, unsigned int cmd) {
struct rb_node * n = guest->vm_ctrls.rb_node;
struct vm_ctrl * ctrl = NULL;
return NULL;
}
+int remove_guest_ctrl(struct v3_guest * guest, unsigned int cmd) {
+ struct vm_ctrl * ctrl = get_ctrl(guest, cmd);
+
+ if (ctrl == NULL) {
+ INFO("Could not find control (%d) to remove\n", cmd);
+ return -1;
+ }
+
+ rb_erase(&(ctrl->tree_node), &(guest->vm_ctrls));
+
+ kfree(ctrl);
+
+ return 0;
+}
+
+static void free_guest_ctrls(struct v3_guest * guest) {
+ struct rb_node * node = rb_first(&(guest->vm_ctrls));
+ struct vm_ctrl * ctrl = NULL;
+ struct rb_node * tmp_node = NULL;
+
+ while (node) {
+ ctrl = rb_entry(node, struct vm_ctrl, tree_node);
+ tmp_node = node;
+ node = rb_next(node);
+
+ WARNING("Cleaning up guest ctrl that was not removed explicitly (%d)\n", ctrl->cmd);
+
+ kfree(ctrl);
+ }
+}
cdev_del(&(guest->cdev));
+ free_guest_ctrls(guest);
+
+
vfree(guest->img);
palacios_free(guest);
void * priv_data),
void * priv_data);
+int remove_guest_ctrl(struct v3_guest * guest, unsigned int cmd);
+
#endif
# define this at the command-line or here
# to get dynamically linked versions
-STATIC = 1
+STATIC = 0
#
# Executables that implement core functionality
#
BASE_EXECS = v3_mem \
v3_mem_free \
- v3_create \
v3_create_bind \
v3_free \
v3_launch \
v3_debug \
v3_send \
v3_receive \
+ v3_pci
+
#
# Examples
v3_cons_sc: v3_cons_sc.c
$(CC) $(CFLAGS) $(CURSES_CFLAGS) $< -lcurses -o $@
+v3_create: v3_create.c ezxml.c
+ $(CC) $(CFLAGS) $^ -o $@
+
#
--- /dev/null
+/* ezxml.c
+ *
+ * Copyright 2004-2006 Aaron Voisine <aaron@voisine.org>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
+ * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <stdarg.h>
+#include <string.h>
+#include <ctype.h>
+#include <unistd.h>
+#include <sys/types.h>
+#ifndef EZXML_NOMMAP
+#include <sys/mman.h>
+#endif // EZXML_NOMMAP
+#include <sys/stat.h>
+#include "ezxml.h"
+
+#define EZXML_WS "\t\r\n " // whitespace
+#define EZXML_ERRL 128 // maximum error string length
+
+typedef struct ezxml_root *ezxml_root_t;
+struct ezxml_root { // additional data for the root tag
+ struct ezxml xml; // is a super-struct built on top of ezxml struct
+ ezxml_t cur; // current xml tree insertion point
+ char *m; // original xml string
+ size_t len; // length of allocated memory for mmap, -1 for malloc
+ char *u; // UTF-8 conversion of string if original was UTF-16
+ char *s; // start of work area
+ char *e; // end of work area
+ char **ent; // general entities (ampersand sequences)
+ char ***attr; // default attributes
+ char ***pi; // processing instructions
+ short standalone; // non-zero if <?xml standalone="yes"?>
+ char err[EZXML_ERRL]; // error string
+};
+
+char *EZXML_NIL[] = { NULL }; // empty, null terminated array of strings
+
+// returns the first child tag with the given name or NULL if not found
+ezxml_t ezxml_child(ezxml_t xml, const char *name)
+{
+ xml = (xml) ? xml->child : NULL;
+ while (xml && strcmp(name, xml->name)) xml = xml->sibling;
+ return xml;
+}
+
+// returns the Nth tag with the same name in the same subsection or NULL if not
+// found
+ezxml_t ezxml_idx(ezxml_t xml, int idx)
+{
+ for (; xml && idx; idx--) xml = xml->next;
+ return xml;
+}
+
+// returns the value of the requested tag attribute or NULL if not found
+const char *ezxml_attr(ezxml_t xml, const char *attr)
+{
+ int i = 0, j = 1;
+ ezxml_root_t root = (ezxml_root_t)xml;
+
+ if (! xml || ! xml->attr) return NULL;
+ while (xml->attr[i] && strcmp(attr, xml->attr[i])) i += 2;
+ if (xml->attr[i]) return xml->attr[i + 1]; // found attribute
+
+ while (root->xml.parent) root = (ezxml_root_t)root->xml.parent; // root tag
+ for (i = 0; root->attr[i] && strcmp(xml->name, root->attr[i][0]); i++);
+ if (! root->attr[i]) return NULL; // no matching default attributes
+ while (root->attr[i][j] && strcmp(attr, root->attr[i][j])) j += 3;
+ return (root->attr[i][j]) ? root->attr[i][j + 1] : NULL; // found default
+}
+
+// same as ezxml_get but takes an already initialized va_list
+ezxml_t ezxml_vget(ezxml_t xml, va_list ap)
+{
+ char *name = va_arg(ap, char *);
+ int idx = -1;
+
+ if (name && *name) {
+ idx = va_arg(ap, int);
+ xml = ezxml_child(xml, name);
+ }
+ return (idx < 0) ? xml : ezxml_vget(ezxml_idx(xml, idx), ap);
+}
+
+// Traverses the xml tree to retrieve a specific subtag. Takes a variable
+// length list of tag names and indexes. The argument list must be terminated
+// by either an index of -1 or an empty string tag name. Example:
+// title = ezxml_get(library, "shelf", 0, "book", 2, "title", -1);
+// This retrieves the title of the 3rd book on the 1st shelf of library.
+// Returns NULL if not found.
+ezxml_t ezxml_get(ezxml_t xml, ...)
+{
+ va_list ap;
+ ezxml_t r;
+
+ va_start(ap, xml);
+ r = ezxml_vget(xml, ap);
+ va_end(ap);
+ return r;
+}
+
+// returns a null terminated array of processing instructions for the given
+// target
+const char **ezxml_pi(ezxml_t xml, const char *target)
+{
+ ezxml_root_t root = (ezxml_root_t)xml;
+ int i = 0;
+
+ if (! root) return (const char **)EZXML_NIL;
+ while (root->xml.parent) root = (ezxml_root_t)root->xml.parent; // root tag
+ while (root->pi[i] && strcmp(target, root->pi[i][0])) i++; // find target
+ return (const char **)((root->pi[i]) ? root->pi[i] + 1 : EZXML_NIL);
+}
+
+// set an error string and return root
+ezxml_t ezxml_err(ezxml_root_t root, char *s, const char *err, ...)
+{
+ va_list ap;
+ int line = 1;
+ char *t, fmt[EZXML_ERRL];
+
+ for (t = root->s; t < s; t++) if (*t == '\n') line++;
+ snprintf(fmt, EZXML_ERRL, "[error near line %d]: %s", line, err);
+
+ va_start(ap, err);
+ vsnprintf(root->err, EZXML_ERRL, fmt, ap);
+ va_end(ap);
+
+ return &root->xml;
+}
+
+// Recursively decodes entity and character references and normalizes new lines
+// ent is a null terminated array of alternating entity names and values. set t
+// to '&' for general entity decoding, '%' for parameter entity decoding, 'c'
+// for cdata sections, ' ' for attribute normalization, or '*' for non-cdata
+// attribute normalization. Returns s, or if the decoded string is longer than
+// s, returns a malloced string that must be freed.
+char *ezxml_decode(char *s, char **ent, char t)
+{
+ char *e, *r = s, *m = s;
+ long b, c, d, l;
+
+ for (; *s; s++) { // normalize line endings
+ while (*s == '\r') {
+ *(s++) = '\n';
+ if (*s == '\n') memmove(s, (s + 1), strlen(s));
+ }
+ }
+
+ for (s = r; ; ) {
+ while (*s && *s != '&' && (*s != '%' || t != '%') && !isspace(*s)) s++;
+
+ if (! *s) break;
+ else if (t != 'c' && ! strncmp(s, "&#", 2)) { // character reference
+ if (s[2] == 'x') c = strtol(s + 3, &e, 16); // base 16
+ else c = strtol(s + 2, &e, 10); // base 10
+ if (! c || *e != ';') { s++; continue; } // not a character ref
+
+ if (c < 0x80) *(s++) = c; // US-ASCII subset
+ else { // multi-byte UTF-8 sequence
+ for (b = 0, d = c; d; d /= 2) b++; // number of bits in c
+ b = (b - 2) / 5; // number of bytes in payload
+ *(s++) = (0xFF << (7 - b)) | (c >> (6 * b)); // head
+ while (b) *(s++) = 0x80 | ((c >> (6 * --b)) & 0x3F); // payload
+ }
+
+ memmove(s, strchr(s, ';') + 1, strlen(strchr(s, ';')));
+ }
+ else if ((*s == '&' && (t == '&' || t == ' ' || t == '*')) ||
+ (*s == '%' && t == '%')) { // entity reference
+ for (b = 0; ent[b] && strncmp(s + 1, ent[b], strlen(ent[b]));
+ b += 2); // find entity in entity list
+
+ if (ent[b++]) { // found a match
+ if ((c = strlen(ent[b])) - 1 > (e = strchr(s, ';')) - s) {
+ l = (d = (s - r)) + c + strlen(e); // new length
+ r = (r == m) ? strcpy(malloc(l), r) : realloc(r, l);
+ e = strchr((s = r + d), ';'); // fix up pointers
+ }
+
+ memmove(s + c, e + 1, strlen(e)); // shift rest of string
+ strncpy(s, ent[b], c); // copy in replacement text
+ }
+ else s++; // not a known entity
+ }
+ else if ((t == ' ' || t == '*') && isspace(*s)) *(s++) = ' ';
+ else s++; // no decoding needed
+ }
+
+ if (t == '*') { // normalize spaces for non-cdata attributes
+ for (s = r; *s; s++) {
+ if ((l = strspn(s, " "))) memmove(s, s + l, strlen(s + l) + 1);
+ while (*s && *s != ' ') s++;
+ }
+ if (--s >= r && *s == ' ') *s = '\0'; // trim any trailing space
+ }
+ return r;
+}
+
+// called when parser finds start of new tag
+void ezxml_open_tag(ezxml_root_t root, char *name, char **attr)
+{
+ ezxml_t xml = root->cur;
+
+ if (xml->name) xml = ezxml_add_child(xml, name, strlen(xml->txt));
+ else xml->name = name; // first open tag
+
+ xml->attr = attr;
+ root->cur = xml; // update tag insertion point
+}
+
+// called when parser finds character content between open and closing tag
+void ezxml_char_content(ezxml_root_t root, char *s, size_t len, char t)
+{
+ ezxml_t xml = root->cur;
+ char *m = s;
+ size_t l;
+
+ if (! xml || ! xml->name || ! len) return; // sanity check
+
+ s[len] = '\0'; // null terminate text (calling functions anticipate this)
+ len = strlen(s = ezxml_decode(s, root->ent, t)) + 1;
+
+ if (! *(xml->txt)) xml->txt = s; // initial character content
+ else { // allocate our own memory and make a copy
+ xml->txt = (xml->flags & EZXML_TXTM) // allocate some space
+ ? realloc(xml->txt, (l = strlen(xml->txt)) + len)
+ : strcpy(malloc((l = strlen(xml->txt)) + len), xml->txt);
+ strcpy(xml->txt + l, s); // add new char content
+ if (s != m) free(s); // free s if it was malloced by ezxml_decode()
+ }
+
+ if (xml->txt != m) ezxml_set_flag(xml, EZXML_TXTM);
+}
+
+// called when parser finds closing tag
+ezxml_t ezxml_close_tag(ezxml_root_t root, char *name, char *s)
+{
+ if (! root->cur || ! root->cur->name || strcmp(name, root->cur->name))
+ return ezxml_err(root, s, "unexpected closing tag </%s>", name);
+
+ root->cur = root->cur->parent;
+ return NULL;
+}
+
+// checks for circular entity references, returns non-zero if no circular
+// references are found, zero otherwise
+int ezxml_ent_ok(char *name, char *s, char **ent)
+{
+ int i;
+
+ for (; ; s++) {
+ while (*s && *s != '&') s++; // find next entity reference
+ if (! *s) return 1;
+ if (! strncmp(s + 1, name, strlen(name))) return 0; // circular ref.
+ for (i = 0; ent[i] && strncmp(ent[i], s + 1, strlen(ent[i])); i += 2);
+ if (ent[i] && ! ezxml_ent_ok(name, ent[i + 1], ent)) return 0;
+ }
+}
+
+// called when the parser finds a processing instruction
+void ezxml_proc_inst(ezxml_root_t root, char *s, size_t len)
+{
+ int i = 0, j = 1;
+ char *target = s;
+
+ s[len] = '\0'; // null terminate instruction
+ if (*(s += strcspn(s, EZXML_WS))) {
+ *s = '\0'; // null terminate target
+ s += strspn(s + 1, EZXML_WS) + 1; // skip whitespace after target
+ }
+
+ if (! strcmp(target, "xml")) { // <?xml ... ?>
+ if ((s = strstr(s, "standalone")) && ! strncmp(s + strspn(s + 10,
+ EZXML_WS "='\"") + 10, "yes", 3)) root->standalone = 1;
+ return;
+ }
+
+ if (! root->pi[0]) *(root->pi = malloc(sizeof(char **))) = NULL; //first pi
+
+ while (root->pi[i] && strcmp(target, root->pi[i][0])) i++; // find target
+ if (! root->pi[i]) { // new target
+ root->pi = realloc(root->pi, sizeof(char **) * (i + 2));
+ root->pi[i] = malloc(sizeof(char *) * 3);
+ root->pi[i][0] = target;
+ root->pi[i][1] = (char *)(root->pi[i + 1] = NULL); // terminate pi list
+ root->pi[i][2] = strdup(""); // empty document position list
+ }
+
+ while (root->pi[i][j]) j++; // find end of instruction list for this target
+ root->pi[i] = realloc(root->pi[i], sizeof(char *) * (j + 3));
+ root->pi[i][j + 2] = realloc(root->pi[i][j + 1], j + 1);
+ strcpy(root->pi[i][j + 2] + j - 1, (root->xml.name) ? ">" : "<");
+ root->pi[i][j + 1] = NULL; // null terminate pi list for this target
+ root->pi[i][j] = s; // set instruction
+}
+
+// called when the parser finds an internal doctype subset
+short ezxml_internal_dtd(ezxml_root_t root, char *s, size_t len)
+{
+ char q, *c, *t, *n = NULL, *v, **ent, **pe;
+ int i, j;
+
+ pe = memcpy(malloc(sizeof(EZXML_NIL)), EZXML_NIL, sizeof(EZXML_NIL));
+
+ for (s[len] = '\0'; s; ) {
+ while (*s && *s != '<' && *s != '%') s++; // find next declaration
+
+ if (! *s) break;
+ else if (! strncmp(s, "<!ENTITY", 8)) { // parse entity definitions
+ c = s += strspn(s + 8, EZXML_WS) + 8; // skip white space separator
+ n = s + strspn(s, EZXML_WS "%"); // find name
+ *(s = n + strcspn(n, EZXML_WS)) = ';'; // append ; to name
+
+ v = s + strspn(s + 1, EZXML_WS) + 1; // find value
+ if ((q = *(v++)) != '"' && q != '\'') { // skip externals
+ s = strchr(s, '>');
+ continue;
+ }
+
+ for (i = 0, ent = (*c == '%') ? pe : root->ent; ent[i]; i++);
+ ent = realloc(ent, (i + 3) * sizeof(char *)); // space for next ent
+ if (*c == '%') pe = ent;
+ else root->ent = ent;
+
+ *(++s) = '\0'; // null terminate name
+ if ((s = strchr(v, q))) *(s++) = '\0'; // null terminate value
+ ent[i + 1] = ezxml_decode(v, pe, '%'); // set value
+ ent[i + 2] = NULL; // null terminate entity list
+ if (! ezxml_ent_ok(n, ent[i + 1], ent)) { // circular reference
+ if (ent[i + 1] != v) free(ent[i + 1]);
+ ezxml_err(root, v, "circular entity declaration &%s", n);
+ break;
+ }
+ else ent[i] = n; // set entity name
+ }
+ else if (! strncmp(s, "<!ATTLIST", 9)) { // parse default attributes
+ t = s + strspn(s + 9, EZXML_WS) + 9; // skip whitespace separator
+ if (! *t) { ezxml_err(root, t, "unclosed <!ATTLIST"); break; }
+ if (*(s = t + strcspn(t, EZXML_WS ">")) == '>') continue;
+ else *s = '\0'; // null terminate tag name
+ for (i = 0; root->attr[i] && strcmp(n, root->attr[i][0]); i++);
+
+ while (*(n = ++s + strspn(s, EZXML_WS)) && *n != '>') {
+ if (*(s = n + strcspn(n, EZXML_WS))) *s = '\0'; // attr name
+ else { ezxml_err(root, t, "malformed <!ATTLIST"); break; }
+
+ s += strspn(s + 1, EZXML_WS) + 1; // find next token
+ c = (strncmp(s, "CDATA", 5)) ? "*" : " "; // is it cdata?
+ if (! strncmp(s, "NOTATION", 8))
+ s += strspn(s + 8, EZXML_WS) + 8;
+ s = (*s == '(') ? strchr(s, ')') : s + strcspn(s, EZXML_WS);
+ if (! s) { ezxml_err(root, t, "malformed <!ATTLIST"); break; }
+
+ s += strspn(s, EZXML_WS ")"); // skip white space separator
+ if (! strncmp(s, "#FIXED", 6))
+ s += strspn(s + 6, EZXML_WS) + 6;
+ if (*s == '#') { // no default value
+ s += strcspn(s, EZXML_WS ">") - 1;
+ if (*c == ' ') continue; // cdata is default, nothing to do
+ v = NULL;
+ }
+ else if ((*s == '"' || *s == '\'') && // default value
+ (s = strchr(v = s + 1, *s))) *s = '\0';
+ else { ezxml_err(root, t, "malformed <!ATTLIST"); break; }
+
+ if (! root->attr[i]) { // new tag name
+ root->attr = (! i) ? malloc(2 * sizeof(char **))
+ : realloc(root->attr,
+ (i + 2) * sizeof(char **));
+ root->attr[i] = malloc(2 * sizeof(char *));
+ root->attr[i][0] = t; // set tag name
+ root->attr[i][1] = (char *)(root->attr[i + 1] = NULL);
+ }
+
+ for (j = 1; root->attr[i][j]; j += 3); // find end of list
+ root->attr[i] = realloc(root->attr[i],
+ (j + 4) * sizeof(char *));
+
+ root->attr[i][j + 3] = NULL; // null terminate list
+ root->attr[i][j + 2] = c; // is it cdata?
+ root->attr[i][j + 1] = (v) ? ezxml_decode(v, root->ent, *c)
+ : NULL;
+ root->attr[i][j] = n; // attribute name
+ }
+ }
+ else if (! strncmp(s, "<!--", 4)) s = strstr(s + 4, "-->"); // comments
+ else if (! strncmp(s, "<?", 2)) { // processing instructions
+ if ((s = strstr(c = s + 2, "?>")))
+ ezxml_proc_inst(root, c, s++ - c);
+ }
+ else if (*s == '<') s = strchr(s, '>'); // skip other declarations
+ else if (*(s++) == '%' && ! root->standalone) break;
+ }
+
+ free(pe);
+ return ! *root->err;
+}
+
+// Converts a UTF-16 string to UTF-8. Returns a new string that must be freed
+// or NULL if no conversion was needed.
+char *ezxml_str2utf8(char **s, size_t *len)
+{
+ char *u;
+ size_t l = 0, sl, max = *len;
+ long c, d;
+ int b, be = (**s == '\xFE') ? 1 : (**s == '\xFF') ? 0 : -1;
+
+ if (be == -1) return NULL; // not UTF-16
+
+ u = malloc(max);
+ for (sl = 2; sl < *len - 1; sl += 2) {
+ c = (be) ? (((*s)[sl] & 0xFF) << 8) | ((*s)[sl + 1] & 0xFF) //UTF-16BE
+ : (((*s)[sl + 1] & 0xFF) << 8) | ((*s)[sl] & 0xFF); //UTF-16LE
+ if (c >= 0xD800 && c <= 0xDFFF && (sl += 2) < *len - 1) { // high-half
+ d = (be) ? (((*s)[sl] & 0xFF) << 8) | ((*s)[sl + 1] & 0xFF)
+ : (((*s)[sl + 1] & 0xFF) << 8) | ((*s)[sl] & 0xFF);
+ c = (((c & 0x3FF) << 10) | (d & 0x3FF)) + 0x10000;
+ }
+
+ while (l + 6 > max) u = realloc(u, max += EZXML_BUFSIZE);
+ if (c < 0x80) u[l++] = c; // US-ASCII subset
+ else { // multi-byte UTF-8 sequence
+ for (b = 0, d = c; d; d /= 2) b++; // bits in c
+ b = (b - 2) / 5; // bytes in payload
+ u[l++] = (0xFF << (7 - b)) | (c >> (6 * b)); // head
+ while (b) u[l++] = 0x80 | ((c >> (6 * --b)) & 0x3F); // payload
+ }
+ }
+ return *s = realloc(u, *len = l);
+}
+
+// frees a tag attribute list
+void ezxml_free_attr(char **attr) {
+ int i = 0;
+ char *m;
+
+ if (! attr || attr == EZXML_NIL) return; // nothing to free
+ while (attr[i]) i += 2; // find end of attribute list
+ m = attr[i + 1]; // list of which names and values are malloced
+ for (i = 0; m[i]; i++) {
+ if (m[i] & EZXML_NAMEM) free(attr[i * 2]);
+ if (m[i] & EZXML_TXTM) free(attr[(i * 2) + 1]);
+ }
+ free(m);
+ free(attr);
+}
+
+// parse the given xml string and return an ezxml structure
+ezxml_t ezxml_parse_str(char *s, size_t len)
+{
+ ezxml_root_t root = (ezxml_root_t)ezxml_new(NULL);
+ char q, e, *d, **attr, **a = NULL; // initialize a to avoid compile warning
+ int l, i, j;
+
+ root->m = s;
+ if (! len) return ezxml_err(root, NULL, "root tag missing");
+ root->u = ezxml_str2utf8(&s, &len); // convert utf-16 to utf-8
+ root->e = (root->s = s) + len; // record start and end of work area
+
+ e = s[len - 1]; // save end char
+ s[len - 1] = '\0'; // turn end char into null terminator
+
+ while (*s && *s != '<') s++; // find first tag
+ if (! *s) return ezxml_err(root, s, "root tag missing");
+
+ for (; ; ) {
+ attr = (char **)EZXML_NIL;
+ d = ++s;
+
+ if (isalpha(*s) || *s == '_' || *s == ':' || *s < '\0') { // new tag
+ if (! root->cur)
+ return ezxml_err(root, d, "markup outside of root element");
+
+ s += strcspn(s, EZXML_WS "/>");
+ while (isspace(*s)) *(s++) = '\0'; // null terminate tag name
+
+ if (*s && *s != '/' && *s != '>') // find tag in default attr list
+ for (i = 0; (a = root->attr[i]) && strcmp(a[0], d); i++);
+
+ for (l = 0; *s && *s != '/' && *s != '>'; l += 2) { // new attrib
+ attr = (l) ? realloc(attr, (l + 4) * sizeof(char *))
+ : malloc(4 * sizeof(char *)); // allocate space
+ attr[l + 3] = (l) ? realloc(attr[l + 1], (l / 2) + 2)
+ : malloc(2); // mem for list of maloced vals
+ strcpy(attr[l + 3] + (l / 2), " "); // value is not malloced
+ attr[l + 2] = NULL; // null terminate list
+ attr[l + 1] = ""; // temporary attribute value
+ attr[l] = s; // set attribute name
+
+ s += strcspn(s, EZXML_WS "=/>");
+ if (*s == '=' || isspace(*s)) {
+ *(s++) = '\0'; // null terminate tag attribute name
+ q = *(s += strspn(s, EZXML_WS "="));
+ if (q == '"' || q == '\'') { // attribute value
+ attr[l + 1] = ++s;
+ while (*s && *s != q) s++;
+ if (*s) *(s++) = '\0'; // null terminate attribute val
+ else {
+ ezxml_free_attr(attr);
+ return ezxml_err(root, d, "missing %c", q);
+ }
+
+ for (j = 1; a && a[j] && strcmp(a[j], attr[l]); j +=3);
+ attr[l + 1] = ezxml_decode(attr[l + 1], root->ent, (a
+ && a[j]) ? *a[j + 2] : ' ');
+ if (attr[l + 1] < d || attr[l + 1] > s)
+ attr[l + 3][l / 2] = EZXML_TXTM; // value malloced
+ }
+ }
+ while (isspace(*s)) s++;
+ }
+
+ if (*s == '/') { // self closing tag
+ *(s++) = '\0';
+ if ((*s && *s != '>') || (! *s && e != '>')) {
+ if (l) ezxml_free_attr(attr);
+ return ezxml_err(root, d, "missing >");
+ }
+ ezxml_open_tag(root, d, attr);
+ ezxml_close_tag(root, d, s);
+ }
+ else if ((q = *s) == '>' || (! *s && e == '>')) { // open tag
+ *s = '\0'; // temporarily null terminate tag name
+ ezxml_open_tag(root, d, attr);
+ *s = q;
+ }
+ else {
+ if (l) ezxml_free_attr(attr);
+ return ezxml_err(root, d, "missing >");
+ }
+ }
+ else if (*s == '/') { // close tag
+ s += strcspn(d = s + 1, EZXML_WS ">") + 1;
+ if (! (q = *s) && e != '>') return ezxml_err(root, d, "missing >");
+ *s = '\0'; // temporarily null terminate tag name
+ if (ezxml_close_tag(root, d, s)) return &root->xml;
+ if (isspace(*s = q)) s += strspn(s, EZXML_WS);
+ }
+ else if (! strncmp(s, "!--", 3)) { // xml comment
+ if (! (s = strstr(s + 3, "--")) || (*(s += 2) != '>' && *s) ||
+ (! *s && e != '>')) return ezxml_err(root, d, "unclosed <!--");
+ }
+ else if (! strncmp(s, "![CDATA[", 8)) { // cdata
+ if ((s = strstr(s, "]]>")))
+ ezxml_char_content(root, d + 8, (s += 2) - d - 10, 'c');
+ else return ezxml_err(root, d, "unclosed <![CDATA[");
+ }
+ else if (! strncmp(s, "!DOCTYPE", 8)) { // dtd
+ for (l = 0; *s && ((! l && *s != '>') || (l && (*s != ']' ||
+ *(s + strspn(s + 1, EZXML_WS) + 1) != '>')));
+ l = (*s == '[') ? 1 : l) s += strcspn(s + 1, "[]>") + 1;
+ if (! *s && e != '>')
+ return ezxml_err(root, d, "unclosed <!DOCTYPE");
+ d = (l) ? strchr(d, '[') + 1 : d;
+ if (l && ! ezxml_internal_dtd(root, d, s++ - d)) return &root->xml;
+ }
+ else if (*s == '?') { // <?...?> processing instructions
+ do { s = strchr(s, '?'); } while (s && *(++s) && *s != '>');
+ if (! s || (! *s && e != '>'))
+ return ezxml_err(root, d, "unclosed <?");
+ else ezxml_proc_inst(root, d + 1, s - d - 2);
+ }
+ else return ezxml_err(root, d, "unexpected <");
+
+ if (! s || ! *s) break;
+ *s = '\0';
+ d = ++s;
+ if (*s && *s != '<') { // tag character content
+ while (*s && *s != '<') s++;
+ if (*s) ezxml_char_content(root, d, s - d, '&');
+ else break;
+ }
+ else if (! *s) break;
+ }
+
+ if (! root->cur) return &root->xml;
+ else if (! root->cur->name) return ezxml_err(root, d, "root tag missing");
+ else return ezxml_err(root, d, "unclosed tag <%s>", root->cur->name);
+}
+
+// Wrapper for ezxml_parse_str() that accepts a file stream. Reads the entire
+// stream into memory and then parses it. For xml files, use ezxml_parse_file()
+// or ezxml_parse_fd()
+ezxml_t ezxml_parse_fp(FILE *fp)
+{
+ ezxml_root_t root;
+ size_t l, len = 0;
+ char *s;
+
+ if (! (s = malloc(EZXML_BUFSIZE))) return NULL;
+ do {
+ len += (l = fread((s + len), 1, EZXML_BUFSIZE, fp));
+ if (l == EZXML_BUFSIZE) s = realloc(s, len + EZXML_BUFSIZE);
+ } while (s && l == EZXML_BUFSIZE);
+
+ if (! s) return NULL;
+ root = (ezxml_root_t)ezxml_parse_str(s, len);
+ root->len = -1; // so we know to free s in ezxml_free()
+ return &root->xml;
+}
+
+// A wrapper for ezxml_parse_str() that accepts a file descriptor. First
+// attempts to mem map the file. Failing that, reads the file into memory.
+// Returns NULL on failure.
+ezxml_t ezxml_parse_fd(int fd)
+{
+ ezxml_root_t root;
+ struct stat st;
+ size_t l;
+ void *m;
+
+ if (fd < 0) return NULL;
+ fstat(fd, &st);
+
+#ifndef EZXML_NOMMAP
+ l = (st.st_size + sysconf(_SC_PAGESIZE) - 1) & ~(sysconf(_SC_PAGESIZE) -1);
+ if ((m = mmap(NULL, l, PROT_READ | PROT_WRITE, MAP_PRIVATE, fd, 0)) !=
+ MAP_FAILED) {
+ madvise(m, l, MADV_SEQUENTIAL); // optimize for sequential access
+ root = (ezxml_root_t)ezxml_parse_str(m, st.st_size);
+ madvise(m, root->len = l, MADV_NORMAL); // put it back to normal
+ }
+ else { // mmap failed, read file into memory
+#endif // EZXML_NOMMAP
+ l = read(fd, m = malloc(st.st_size), st.st_size);
+ root = (ezxml_root_t)ezxml_parse_str(m, l);
+ root->len = -1; // so we know to free s in ezxml_free()
+#ifndef EZXML_NOMMAP
+ }
+#endif // EZXML_NOMMAP
+ return &root->xml;
+}
+
+// a wrapper for ezxml_parse_fd that accepts a file name
+ezxml_t ezxml_parse_file(const char *file)
+{
+ int fd = open(file, O_RDONLY, 0);
+ ezxml_t xml = ezxml_parse_fd(fd);
+
+ if (fd >= 0) close(fd);
+ return xml;
+}
+
+// Encodes ampersand sequences appending the results to *dst, reallocating *dst
+// if length excedes max. a is non-zero for attribute encoding. Returns *dst
+char *ezxml_ampencode(const char *s, size_t len, char **dst, size_t *dlen,
+ size_t *max, short a)
+{
+ const char *e;
+
+ for (e = s + len; s != e; s++) {
+ while (*dlen + 10 > *max) *dst = realloc(*dst, *max += EZXML_BUFSIZE);
+
+ switch (*s) {
+ case '\0': return *dst;
+ case '&': *dlen += sprintf(*dst + *dlen, "&"); break;
+ case '<': *dlen += sprintf(*dst + *dlen, "<"); break;
+ case '>': *dlen += sprintf(*dst + *dlen, ">"); break;
+ case '"': *dlen += sprintf(*dst + *dlen, (a) ? """ : "\""); break;
+ case '\n': *dlen += sprintf(*dst + *dlen, (a) ? "
" : "\n"); break;
+ case '\t': *dlen += sprintf(*dst + *dlen, (a) ? "	" : "\t"); break;
+ case '\r': *dlen += sprintf(*dst + *dlen, "
"); break;
+ default: (*dst)[(*dlen)++] = *s;
+ }
+ }
+ return *dst;
+}
+
+// Recursively converts each tag to xml appending it to *s. Reallocates *s if
+// its length excedes max. start is the location of the previous tag in the
+// parent tag's character content. Returns *s.
+char *ezxml_toxml_r(ezxml_t xml, char **s, size_t *len, size_t *max,
+ size_t start, char ***attr)
+{
+ int i, j;
+ char *txt = (xml->parent) ? xml->parent->txt : "";
+ size_t off = 0;
+
+ // parent character content up to this tag
+ *s = ezxml_ampencode(txt + start, xml->off - start, s, len, max, 0);
+
+ while (*len + strlen(xml->name) + 4 > *max) // reallocate s
+ *s = realloc(*s, *max += EZXML_BUFSIZE);
+
+ *len += sprintf(*s + *len, "<%s", xml->name); // open tag
+ for (i = 0; xml->attr[i]; i += 2) { // tag attributes
+ if (ezxml_attr(xml, xml->attr[i]) != xml->attr[i + 1]) continue;
+ while (*len + strlen(xml->attr[i]) + 7 > *max) // reallocate s
+ *s = realloc(*s, *max += EZXML_BUFSIZE);
+
+ *len += sprintf(*s + *len, " %s=\"", xml->attr[i]);
+ ezxml_ampencode(xml->attr[i + 1], -1, s, len, max, 1);
+ *len += sprintf(*s + *len, "\"");
+ }
+
+ for (i = 0; attr[i] && strcmp(attr[i][0], xml->name); i++);
+ for (j = 1; attr[i] && attr[i][j]; j += 3) { // default attributes
+ if (! attr[i][j + 1] || ezxml_attr(xml, attr[i][j]) != attr[i][j + 1])
+ continue; // skip duplicates and non-values
+ while (*len + strlen(attr[i][j]) + 7 > *max) // reallocate s
+ *s = realloc(*s, *max += EZXML_BUFSIZE);
+
+ *len += sprintf(*s + *len, " %s=\"", attr[i][j]);
+ ezxml_ampencode(attr[i][j + 1], -1, s, len, max, 1);
+ *len += sprintf(*s + *len, "\"");
+ }
+ *len += sprintf(*s + *len, ">");
+
+ *s = (xml->child) ? ezxml_toxml_r(xml->child, s, len, max, 0, attr) //child
+ : ezxml_ampencode(xml->txt, -1, s, len, max, 0); //data
+
+ while (*len + strlen(xml->name) + 4 > *max) // reallocate s
+ *s = realloc(*s, *max += EZXML_BUFSIZE);
+
+ *len += sprintf(*s + *len, "</%s>", xml->name); // close tag
+
+ while (txt[off] && off < xml->off) off++; // make sure off is within bounds
+ return (xml->ordered) ? ezxml_toxml_r(xml->ordered, s, len, max, off, attr)
+ : ezxml_ampencode(txt + off, -1, s, len, max, 0);
+}
+
+// Converts an ezxml structure back to xml. Returns a string of xml data that
+// must be freed.
+char *ezxml_toxml(ezxml_t xml)
+{
+ ezxml_t p = (xml) ? xml->parent : NULL, o = (xml) ? xml->ordered : NULL;
+ ezxml_root_t root = (ezxml_root_t)xml;
+ size_t len = 0, max = EZXML_BUFSIZE;
+ char *s = strcpy(malloc(max), ""), *t, *n;
+ int i, j, k;
+
+ if (! xml || ! xml->name) return realloc(s, len + 1);
+ while (root->xml.parent) root = (ezxml_root_t)root->xml.parent; // root tag
+
+ for (i = 0; ! p && root->pi[i]; i++) { // pre-root processing instructions
+ for (k = 2; root->pi[i][k - 1]; k++);
+ for (j = 1; (n = root->pi[i][j]); j++) {
+ if (root->pi[i][k][j - 1] == '>') continue; // not pre-root
+ while (len + strlen(t = root->pi[i][0]) + strlen(n) + 7 > max)
+ s = realloc(s, max += EZXML_BUFSIZE);
+ len += sprintf(s + len, "<?%s%s%s?>\n", t, *n ? " " : "", n);
+ }
+ }
+
+ xml->parent = xml->ordered = NULL;
+ s = ezxml_toxml_r(xml, &s, &len, &max, 0, root->attr);
+ xml->parent = p;
+ xml->ordered = o;
+
+ for (i = 0; ! p && root->pi[i]; i++) { // post-root processing instructions
+ for (k = 2; root->pi[i][k - 1]; k++);
+ for (j = 1; (n = root->pi[i][j]); j++) {
+ if (root->pi[i][k][j - 1] == '<') continue; // not post-root
+ while (len + strlen(t = root->pi[i][0]) + strlen(n) + 7 > max)
+ s = realloc(s, max += EZXML_BUFSIZE);
+ len += sprintf(s + len, "\n<?%s%s%s?>", t, *n ? " " : "", n);
+ }
+ }
+ return realloc(s, len + 1);
+}
+
+// free the memory allocated for the ezxml structure
+void ezxml_free(ezxml_t xml)
+{
+ ezxml_root_t root = (ezxml_root_t)xml;
+ int i, j;
+ char **a, *s;
+
+ if (! xml) return;
+ ezxml_free(xml->child);
+ ezxml_free(xml->ordered);
+
+ if (! xml->parent) { // free root tag allocations
+ for (i = 10; root->ent[i]; i += 2) // 0 - 9 are default entites (<>&"')
+ if ((s = root->ent[i + 1]) < root->s || s > root->e) free(s);
+ free(root->ent); // free list of general entities
+
+ for (i = 0; (a = root->attr[i]); i++) {
+ for (j = 1; a[j++]; j += 2) // free malloced attribute values
+ if (a[j] && (a[j] < root->s || a[j] > root->e)) free(a[j]);
+ free(a);
+ }
+ if (root->attr[0]) free(root->attr); // free default attribute list
+
+ for (i = 0; root->pi[i]; i++) {
+ for (j = 1; root->pi[i][j]; j++);
+ free(root->pi[i][j + 1]);
+ free(root->pi[i]);
+ }
+ if (root->pi[0]) free(root->pi); // free processing instructions
+
+ if (root->len == -1) free(root->m); // malloced xml data
+#ifndef EZXML_NOMMAP
+ else if (root->len) munmap(root->m, root->len); // mem mapped xml data
+#endif // EZXML_NOMMAP
+ if (root->u) free(root->u); // utf8 conversion
+ }
+
+ ezxml_free_attr(xml->attr); // tag attributes
+ if ((xml->flags & EZXML_TXTM)) free(xml->txt); // character content
+ if ((xml->flags & EZXML_NAMEM)) free(xml->name); // tag name
+ free(xml);
+}
+
+// return parser error message or empty string if none
+const char *ezxml_error(ezxml_t xml)
+{
+ while (xml && xml->parent) xml = xml->parent; // find root tag
+ return (xml) ? ((ezxml_root_t)xml)->err : "";
+}
+
+// returns a new empty ezxml structure with the given root tag name
+ezxml_t ezxml_new(const char *name)
+{
+ static char *ent[] = { "lt;", "<", "gt;", ">", "quot;", """,
+ "apos;", "'", "amp;", "&", NULL };
+ ezxml_root_t root = (ezxml_root_t)memset(malloc(sizeof(struct ezxml_root)),
+ '\0', sizeof(struct ezxml_root));
+ root->xml.name = (char *)name;
+ root->cur = &root->xml;
+ strcpy(root->err, root->xml.txt = "");
+ root->ent = memcpy(malloc(sizeof(ent)), ent, sizeof(ent));
+ root->attr = root->pi = (char ***)(root->xml.attr = EZXML_NIL);
+ return &root->xml;
+}
+
+// inserts an existing tag into an ezxml structure
+ezxml_t ezxml_insert(ezxml_t xml, ezxml_t dest, size_t off)
+{
+ ezxml_t cur, prev, head;
+
+ xml->next = xml->sibling = xml->ordered = NULL;
+ xml->off = off;
+ xml->parent = dest;
+
+ if ((head = dest->child)) { // already have sub tags
+ if (head->off <= off) { // not first subtag
+ for (cur = head; cur->ordered && cur->ordered->off <= off;
+ cur = cur->ordered);
+ xml->ordered = cur->ordered;
+ cur->ordered = xml;
+ }
+ else { // first subtag
+ xml->ordered = head;
+ dest->child = xml;
+ }
+
+ for (cur = head, prev = NULL; cur && strcmp(cur->name, xml->name);
+ prev = cur, cur = cur->sibling); // find tag type
+ if (cur && cur->off <= off) { // not first of type
+ while (cur->next && cur->next->off <= off) cur = cur->next;
+ xml->next = cur->next;
+ cur->next = xml;
+ }
+ else { // first tag of this type
+ if (prev && cur) prev->sibling = cur->sibling; // remove old first
+ xml->next = cur; // old first tag is now next
+ for (cur = head, prev = NULL; cur && cur->off <= off;
+ prev = cur, cur = cur->sibling); // new sibling insert point
+ xml->sibling = cur;
+ if (prev) prev->sibling = xml;
+ }
+ }
+ else dest->child = xml; // only sub tag
+
+ return xml;
+}
+
+// Adds a child tag. off is the offset of the child tag relative to the start
+// of the parent tag's character content. Returns the child tag.
+ezxml_t ezxml_add_child(ezxml_t xml, const char *name, size_t off)
+{
+ ezxml_t child;
+
+ if (! xml) return NULL;
+ child = (ezxml_t)memset(malloc(sizeof(struct ezxml)), '\0',
+ sizeof(struct ezxml));
+ child->name = (char *)name;
+ child->attr = EZXML_NIL;
+ child->txt = "";
+
+ return ezxml_insert(child, xml, off);
+}
+
+// sets the character content for the given tag and returns the tag
+ezxml_t ezxml_set_txt(ezxml_t xml, const char *txt)
+{
+ if (! xml) return NULL;
+ if (xml->flags & EZXML_TXTM) free(xml->txt); // existing txt was malloced
+ xml->flags &= ~EZXML_TXTM;
+ xml->txt = (char *)txt;
+ return xml;
+}
+
+// Sets the given tag attribute or adds a new attribute if not found. A value
+// of NULL will remove the specified attribute. Returns the tag given.
+ezxml_t ezxml_set_attr(ezxml_t xml, const char *name, const char *value)
+{
+ int l = 0, c;
+
+ if (! xml) return NULL;
+ while (xml->attr[l] && strcmp(xml->attr[l], name)) l += 2;
+ if (! xml->attr[l]) { // not found, add as new attribute
+ if (! value) return xml; // nothing to do
+ if (xml->attr == EZXML_NIL) { // first attribute
+ xml->attr = malloc(4 * sizeof(char *));
+ xml->attr[1] = strdup(""); // empty list of malloced names/vals
+ }
+ else xml->attr = realloc(xml->attr, (l + 4) * sizeof(char *));
+
+ xml->attr[l] = (char *)name; // set attribute name
+ xml->attr[l + 2] = NULL; // null terminate attribute list
+ xml->attr[l + 3] = realloc(xml->attr[l + 1],
+ (c = strlen(xml->attr[l + 1])) + 2);
+ strcpy(xml->attr[l + 3] + c, " "); // set name/value as not malloced
+ if (xml->flags & EZXML_DUP) xml->attr[l + 3][c] = EZXML_NAMEM;
+ }
+ else if (xml->flags & EZXML_DUP) free((char *)name); // name was strduped
+
+ for (c = l; xml->attr[c]; c += 2); // find end of attribute list
+ if (xml->attr[c + 1][l / 2] & EZXML_TXTM) free(xml->attr[l + 1]); //old val
+ if (xml->flags & EZXML_DUP) xml->attr[c + 1][l / 2] |= EZXML_TXTM;
+ else xml->attr[c + 1][l / 2] &= ~EZXML_TXTM;
+
+ if (value) xml->attr[l + 1] = (char *)value; // set attribute value
+ else { // remove attribute
+ if (xml->attr[c + 1][l / 2] & EZXML_NAMEM) free(xml->attr[l]);
+ memmove(xml->attr + l, xml->attr + l + 2, (c - l + 2) * sizeof(char*));
+ xml->attr = realloc(xml->attr, (c + 2) * sizeof(char *));
+ memmove(xml->attr[c + 1] + (l / 2), xml->attr[c + 1] + (l / 2) + 1,
+ (c / 2) - (l / 2)); // fix list of which name/vals are malloced
+ }
+ xml->flags &= ~EZXML_DUP; // clear strdup() flag
+ return xml;
+}
+
+// sets a flag for the given tag and returns the tag
+ezxml_t ezxml_set_flag(ezxml_t xml, short flag)
+{
+ if (xml) xml->flags |= flag;
+ return xml;
+}
+
+// removes a tag along with its subtags without freeing its memory
+ezxml_t ezxml_cut(ezxml_t xml)
+{
+ ezxml_t cur;
+
+ if (! xml) return NULL; // nothing to do
+ if (xml->next) xml->next->sibling = xml->sibling; // patch sibling list
+
+ if (xml->parent) { // not root tag
+ cur = xml->parent->child; // find head of subtag list
+ if (cur == xml) xml->parent->child = xml->ordered; // first subtag
+ else { // not first subtag
+ while (cur->ordered != xml) cur = cur->ordered;
+ cur->ordered = cur->ordered->ordered; // patch ordered list
+
+ cur = xml->parent->child; // go back to head of subtag list
+ if (strcmp(cur->name, xml->name)) { // not in first sibling list
+ while (strcmp(cur->sibling->name, xml->name))
+ cur = cur->sibling;
+ if (cur->sibling == xml) { // first of a sibling list
+ cur->sibling = (xml->next) ? xml->next
+ : cur->sibling->sibling;
+ }
+ else cur = cur->sibling; // not first of a sibling list
+ }
+
+ while (cur->next && cur->next != xml) cur = cur->next;
+ if (cur->next) cur->next = cur->next->next; // patch next list
+ }
+ }
+ xml->ordered = xml->sibling = xml->next = NULL;
+ return xml;
+}
+
+#ifdef EZXML_TEST // test harness
+int main(int argc, char **argv)
+{
+ ezxml_t xml;
+ char *s;
+ int i;
+
+ if (argc != 2) return fprintf(stderr, "usage: %s xmlfile\n", argv[0]);
+
+ xml = ezxml_parse_file(argv[1]);
+ printf("%s\n", (s = ezxml_toxml(xml)));
+ free(s);
+ i = fprintf(stderr, "%s", ezxml_error(xml));
+ ezxml_free(xml);
+ return (i) ? 1 : 0;
+}
+#endif // EZXML_TEST
--- /dev/null
+/* ezxml.h
+ *
+ * Copyright 2004-2006 Aaron Voisine <aaron@voisine.org>
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining
+ * a copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sublicense, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
+ * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef _EZXML_H
+#define _EZXML_H
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <stdarg.h>
+#include <fcntl.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define EZXML_BUFSIZE 1024 // size of internal memory buffers
+#define EZXML_NAMEM 0x80 // name is malloced
+#define EZXML_TXTM 0x40 // txt is malloced
+#define EZXML_DUP 0x20 // attribute name and value are strduped
+
+typedef struct ezxml *ezxml_t;
+struct ezxml {
+ char *name; // tag name
+ char **attr; // tag attributes { name, value, name, value, ... NULL }
+ char *txt; // tag character content, empty string if none
+ size_t off; // tag offset from start of parent tag character content
+ ezxml_t next; // next tag with same name in this section at this depth
+ ezxml_t sibling; // next tag with different name in same section and depth
+ ezxml_t ordered; // next tag, same section and depth, in original order
+ ezxml_t child; // head of sub tag list, NULL if none
+ ezxml_t parent; // parent tag, NULL if current tag is root tag
+ short flags; // additional information
+};
+
+// Given a string of xml data and its length, parses it and creates an ezxml
+// structure. For efficiency, modifies the data by adding null terminators
+// and decoding ampersand sequences. If you don't want this, copy the data and
+// pass in the copy. Returns NULL on failure.
+ezxml_t ezxml_parse_str(char *s, size_t len);
+
+// A wrapper for ezxml_parse_str() that accepts a file descriptor. First
+// attempts to mem map the file. Failing that, reads the file into memory.
+// Returns NULL on failure.
+ezxml_t ezxml_parse_fd(int fd);
+
+// a wrapper for ezxml_parse_fd() that accepts a file name
+ezxml_t ezxml_parse_file(const char *file);
+
+// Wrapper for ezxml_parse_str() that accepts a file stream. Reads the entire
+// stream into memory and then parses it. For xml files, use ezxml_parse_file()
+// or ezxml_parse_fd()
+ezxml_t ezxml_parse_fp(FILE *fp);
+
+// returns the first child tag (one level deeper) with the given name or NULL
+// if not found
+ezxml_t ezxml_child(ezxml_t xml, const char *name);
+
+// returns the next tag of the same name in the same section and depth or NULL
+// if not found
+#define ezxml_next(xml) ((xml) ? xml->next : NULL)
+
+// Returns the Nth tag with the same name in the same section at the same depth
+// or NULL if not found. An index of 0 returns the tag given.
+ezxml_t ezxml_idx(ezxml_t xml, int idx);
+
+// returns the name of the given tag
+#define ezxml_name(xml) ((xml) ? xml->name : NULL)
+
+// returns the given tag's character content or empty string if none
+#define ezxml_txt(xml) ((xml) ? xml->txt : "")
+
+// returns the value of the requested tag attribute, or NULL if not found
+const char *ezxml_attr(ezxml_t xml, const char *attr);
+
+// Traverses the ezxml sturcture to retrieve a specific subtag. Takes a
+// variable length list of tag names and indexes. The argument list must be
+// terminated by either an index of -1 or an empty string tag name. Example:
+// title = ezxml_get(library, "shelf", 0, "book", 2, "title", -1);
+// This retrieves the title of the 3rd book on the 1st shelf of library.
+// Returns NULL if not found.
+ezxml_t ezxml_get(ezxml_t xml, ...);
+
+// Converts an ezxml structure back to xml. Returns a string of xml data that
+// must be freed.
+char *ezxml_toxml(ezxml_t xml);
+
+// returns a NULL terminated array of processing instructions for the given
+// target
+const char **ezxml_pi(ezxml_t xml, const char *target);
+
+// frees the memory allocated for an ezxml structure
+void ezxml_free(ezxml_t xml);
+
+// returns parser error message or empty string if none
+const char *ezxml_error(ezxml_t xml);
+
+// returns a new empty ezxml structure with the given root tag name
+ezxml_t ezxml_new(const char *name);
+
+// wrapper for ezxml_new() that strdup()s name
+#define ezxml_new_d(name) ezxml_set_flag(ezxml_new(strdup(name)), EZXML_NAMEM)
+
+// Adds a child tag. off is the offset of the child tag relative to the start
+// of the parent tag's character content. Returns the child tag.
+ezxml_t ezxml_add_child(ezxml_t xml, const char *name, size_t off);
+
+// wrapper for ezxml_add_child() that strdup()s name
+#define ezxml_add_child_d(xml, name, off) \
+ ezxml_set_flag(ezxml_add_child(xml, strdup(name), off), EZXML_NAMEM)
+
+// sets the character content for the given tag and returns the tag
+ezxml_t ezxml_set_txt(ezxml_t xml, const char *txt);
+
+// wrapper for ezxml_set_txt() that strdup()s txt
+#define ezxml_set_txt_d(xml, txt) \
+ ezxml_set_flag(ezxml_set_txt(xml, strdup(txt)), EZXML_TXTM)
+
+// Sets the given tag attribute or adds a new attribute if not found. A value
+// of NULL will remove the specified attribute. Returns the tag given.
+ezxml_t ezxml_set_attr(ezxml_t xml, const char *name, const char *value);
+
+// Wrapper for ezxml_set_attr() that strdup()s name/value. Value cannot be NULL
+#define ezxml_set_attr_d(xml, name, value) \
+ ezxml_set_attr(ezxml_set_flag(xml, EZXML_DUP), strdup(name), strdup(value))
+
+// sets a flag for the given tag and returns the tag
+ezxml_t ezxml_set_flag(ezxml_t xml, short flag);
+
+// removes a tag along with its subtags without freeing its memory
+ezxml_t ezxml_cut(ezxml_t xml);
+
+// inserts an existing tag into an ezxml structure
+ezxml_t ezxml_insert(ezxml_t xml, ezxml_t dest, size_t off);
+
+// Moves an existing tag to become a subtag of dest at the given offset from
+// the start of dest's character content. Returns the moved tag.
+#define ezxml_move(xml, dest, off) ezxml_insert(ezxml_cut(xml), dest, off)
+
+// removes a tag along with all its subtags
+#define ezxml_remove(xml) ezxml_free(ezxml_cut(xml))
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif // _EZXML_H
* (c) Jack lange, 2010
*/
-
#include <stdio.h>
#include <stdlib.h>
#include <fcntl.h>
#include <sys/types.h>
#include <unistd.h>
#include <string.h>
-
+#include <getopt.h>
+
+
#include "v3_ctrl.h"
+#include "ezxml.h"
+
+struct cfg_value {
+ char * tag;
+ char * value;
+};
+
+struct xml_option {
+ char * tag;
+ ezxml_t location;
+ struct xml_option * next;
+};
+
+
+struct file_info {
+ int size;
+ char filename[2048];
+ char id[256];
+};
+
+#define MAX_FILES 256
+unsigned long long num_files = 0;
+struct file_info files[MAX_FILES];
+
int read_file(int fd, int size, unsigned char * buf);
-int main(int argc, char* argv[]) {
- char * filename = argv[1];
- char * name = argv[2];
- int guest_fd = 0;
- int v3_fd = 0;
+
+
+int create_vm(char * vm_name, void * img_data, unsigned int img_size) {
struct v3_guest_img guest_img;
- struct stat guest_stats;
+ int v3_fd = 0;
int dev_idx = 0;
memset(&guest_img, 0, sizeof(struct v3_guest_img));
- if (argc <= 2) {
- printf("usage: v3_create <guest_img> <vm name>\n");
+ guest_img.size = img_size;
+ guest_img.guest_data = img_data;
+ strncpy(guest_img.name, vm_name, 127);
+
+
+ v3_fd = open(v3_dev, O_RDONLY);
+
+ if (v3_fd == -1) {
+ printf("Error opening V3Vee control device\n");
return -1;
}
- printf("Creating guest: %s\n", filename);
+ dev_idx = ioctl(v3_fd, V3_CREATE_GUEST, &guest_img);
+
+
+ if (dev_idx < 0) {
+ printf("Error (%d) creating VM\n", dev_idx);
+ return -1;
+ }
+
+ printf("VM (%s) created at /dev/v3-vm%d\n", vm_name, dev_idx);
+
+ /* Close the file descriptor. */
+ close(v3_fd);
+
+ return 0;
+}
+
+
+int load_image(char * vm_name, char * filename) {
+ int guest_fd = 0;
+ struct stat guest_stats;
+ int img_size = 0;
+ void * img_data = NULL;
guest_fd = open(filename, O_RDONLY);
printf("ERROR: Could not stat guest image file -- %s\n", filename);
return -1;
}
+
+ img_size = guest_stats.st_size;
+
+ // load guest image into user memory
+ img_data = malloc(img_size);
+ read_file(guest_fd, img_size, img_data);
- guest_img.size = guest_stats.st_size;
+ close(guest_fd);
+
+ printf("Guest image Loaded (size=%u)\n", img_size);
+ return create_vm(vm_name, img_data, img_size);
+
+}
+
+
+ezxml_t open_xml_file(char * filename) {
+
+ ezxml_t xml_input = ezxml_parse_file(filename);
- // load guest image into user memory
- guest_img.guest_data = malloc(guest_img.size);
- if (!guest_img.guest_data) {
- printf("ERROR: could not allocate memory for guest image\n");
- return -1;
+ if (xml_input == NULL) {
+ printf("Error: Could not open XML input file (%s)\n", filename);
+ return NULL;
+ } else if (strcmp("", ezxml_error(xml_input)) != 0) {
+ printf("%s\n", ezxml_error(xml_input));
+ return NULL;
}
- read_file(guest_fd, guest_img.size, guest_img.guest_data);
+ return xml_input;
+}
+
+
+int find_xml_options(ezxml_t xml, struct xml_option ** opts) {
+ int num_opts = 0;
+ ezxml_t child = xml->child;
+ struct xml_option * next_opt = NULL;
+
+ char * opt = (char *)ezxml_attr(xml, "opt_tag");
+
+ if (opt) {
+ next_opt = malloc(sizeof(struct xml_option));
+
+ memset(next_opt, 0, sizeof(struct xml_option));
+
+ next_opt->tag = opt;
+ next_opt->location = xml;
+ next_opt->next = NULL;
+
+// printf("Option found: %s\n", opt);
+
+ *opts = next_opt;
+ num_opts++;
+ }
+
+
+ while (child) {
+
+ fflush(stdout);
+
+ if (next_opt != 0x0) {
+ num_opts += find_xml_options(child, &(next_opt->next));
+ } else {
+ num_opts += find_xml_options(child, opts);
+
+ if (*opts) {
+ next_opt = *opts;
+ }
+ }
+
+ if (next_opt) {
+ while (next_opt->next) {
+ next_opt = next_opt->next;
+ }
+ }
+
+ child = child->ordered;
+ }
- close(guest_fd);
+ return num_opts;
+
+}
+
+
+char * get_val(ezxml_t cfg, char * tag) {
+ char * attrib = (char *)ezxml_attr(cfg, tag);
+ ezxml_t txt = ezxml_child(cfg, tag);
+
+ if ((txt != NULL) && (attrib != NULL)) {
+ printf("Invalid Cfg file: Duplicate value for %s (attr=%s, txt=%s)\n",
+ tag, attrib, ezxml_txt(txt));
+ exit(-1);
+ }
+
+ return (attrib == NULL) ? ezxml_txt(txt) : attrib;
+}
+
+
+int parse_aux_files(ezxml_t cfg_input) {
+ ezxml_t file_tags = NULL;
+ ezxml_t tmp_file_tag = NULL;
+
+ // files are transformed into blobs that are slapped to the end of the file
+
+ file_tags = ezxml_child(cfg_input, "files");
- printf("Loaded guest image. Creation begins.\n");
+ tmp_file_tag = ezxml_child(file_tags, "file");
+
+ while (tmp_file_tag) {
+ char * filename = get_val(tmp_file_tag, "filename");
+ struct stat file_stats;
+ char * id = get_val(tmp_file_tag, "id");
+ char index_buf[256];
+
+
+ if (stat(filename, &file_stats) != 0) {
+ perror(filename);
+ exit(-1);
+ }
+
+ files[num_files].size = (unsigned int)file_stats.st_size;
+ strncpy(files[num_files].id, id, 256);
+ strncpy(files[num_files].filename, filename, 2048);
+
+ snprintf(index_buf, 256, "%llu", num_files);
+ ezxml_set_attr_d(tmp_file_tag, "index", index_buf);
+
+ num_files++;
+ tmp_file_tag = ezxml_next(tmp_file_tag);
+ }
+
+
+ return 0;
+}
+
+int build_image(char * vm_name, char * filename, struct cfg_value * cfg_vals, int num_options) {
+ int i = 0;
+ ezxml_t xml = NULL;
+ struct xml_option * xml_opts = NULL;
+ int num_xml_opts = 0;
+ void * guest_img_data = NULL;
+ int guest_img_size = 0;
+
+
+ xml = open_xml_file(filename);
+
+ // parse options
+ num_xml_opts = find_xml_options(xml, &xml_opts);
- strncpy(guest_img.name, name, 127);
+ // printf("%d options\n", num_xml_opts);
+ // apply options
+ for (i = 0; i < num_options; i++) {
+ struct cfg_value * cfg_val = &cfg_vals[i];
+ struct xml_option * xml_opt = xml_opts;
- v3_fd = open(v3_dev, O_RDONLY);
+ while (xml_opt) {
+ if (strcasecmp(cfg_val->tag, xml_opt->tag) == 0) {
+ break;
+ }
+
+ xml_opt = xml_opt->next;
+ }
- if (v3_fd == -1) {
- printf("Error opening V3Vee control device\n");
- return -1;
+
+ if (!xml_opt) {
+ printf("Could not find Config option (%s) in XML file\n", cfg_val->tag);
+ return -1;
+ }
+
+ ezxml_set_txt(xml_opt->location, cfg_val->value);
}
+
- dev_idx = ioctl(v3_fd, V3_CREATE_GUEST, &guest_img);
+ // parse files
+ parse_aux_files(xml);
+
+ // create image data blob
+ {
+ char * new_xml_str = ezxml_toxml(xml);
+ int file_data_size = 0;
+ int i = 0;
+ int offset = 0;
+ unsigned long long file_offset = 0;
- if (dev_idx < 0) {
- printf("Error (%d) creating VM\n", dev_idx);
+ /* Image size is:
+ 8 byte header +
+ 4 byte xml length +
+ xml strlen +
+ 8 bytes of zeros +
+ 8 bytes (number of files) +
+ num_files * 16 byte file header +
+ 8 bytes of zeroes +
+ file data
+ */
+ for (i = 0; i < num_files; i++) {
+ file_data_size += files[i].size;
+ }
+
+ guest_img_size = 8 + 4 + strlen(new_xml_str) + 8 + 8 +
+ (num_files * 16) + 8 + file_data_size;
+
+
+ guest_img_data = malloc(guest_img_size);
+ memset(guest_img_data, 0, guest_img_size);
+
+ memcpy(guest_img_data, "v3vee\0\0\0", 8);
+ offset += 8;
+
+ *(unsigned int *)(guest_img_data + offset) = strlen(new_xml_str);
+ offset += 4;
+
+ memcpy(guest_img_data + offset, new_xml_str, strlen(new_xml_str));
+ offset += strlen(new_xml_str);
+
+ memset(guest_img_data + offset, 0, 8);
+ offset += 8;
+
+ *(unsigned long long *)(guest_img_data + offset) = num_files;
+ offset += 8;
+
+
+ // The file offset starts at the end of the file list
+ file_offset = offset + (16 * num_files) + 8;
+
+ for (i = 0; i < num_files; i++) {
+ *(unsigned int *)(guest_img_data + offset) = i;
+ offset += 4;
+ *(unsigned int *)(guest_img_data + offset) = files[i].size;
+ offset += 4;
+ *(unsigned long long *)(guest_img_data + offset) = file_offset;
+ offset += 8;
+
+ file_offset += files[i].size;
+
+ }
+
+ memset(guest_img_data + offset, 0, 8);
+ offset += 8;
+
+
+ for (i = 0; i < num_files; i++) {
+ int fd = open(files[i].filename, O_RDONLY);
+
+ if (fd == -1) {
+ printf("Error: Could not open aux file (%s)\n", files[i].filename);
+ return -1;
+ }
+
+ read_file(fd, files[i].size, (unsigned char *)(guest_img_data + offset));
+
+ close(fd);
+
+ offset += files[i].size;
+
+ }
+
+ free(new_xml_str);
+
+ }
+
+ printf("Guest Image Created (size=%u)\n", guest_img_size);
+ return create_vm(vm_name, guest_img_data, guest_img_size);
+}
+
+
+
+
+
+
+int main(int argc, char** argv) {
+ char * filename = NULL;
+ char * name = NULL;
+ int build_flag = 0;
+ int c = 0;
+
+ opterr = 0;
+
+ while (( c = getopt(argc, argv, "b")) != -1) {
+ switch (c) {
+ case 'b':
+ build_flag = 1;
+ break;
+ }
+ }
+
+ if (argc - optind + 1 < 3) {
+ printf("usage: v3_create [-b] <guest_img> <vm name> [cfg options]\n");
return -1;
}
- printf("VM (%s) created at /dev/v3-vm%d\n", name, dev_idx);
+ filename = argv[optind];
+ name = argv[optind + 1];
- /* Close the file descriptor. */
- close(v3_fd);
-
+ if (build_flag == 1) {
+ int cfg_idx = optind + 2;
+ int i = 0;
+ struct cfg_value * cfg_vals = NULL;
+
+ printf("Building VM Image (cfg=%s) (name=%s) (%d config options)\n", filename, name, argc - cfg_idx);
+
+ cfg_vals = malloc(sizeof(struct cfg_value) * argc - cfg_idx);
+
+
+
+ while (i < argc - cfg_idx) {
+ char * tag = NULL;
+ char * value = NULL;
+
+ value = argv[cfg_idx + i];
+
+ tag = strsep(&value, "=");
+ // parse
+
+ if (value == NULL) {
+ printf("Invalid Option format: %s\n", argv[cfg_idx + i]);
+ return -1;
+ }
+
+ cfg_vals[i].tag = tag;
+ cfg_vals[i].value = value;
+
+ printf("Config option: %s: %s\n", tag, value);
+
+
+ i++;
+ }
+
+ return build_image(name, filename, cfg_vals, argc - cfg_idx);
+
+
+ } else {
+ printf("Loading VM Image (img=%s) (name=%s)\n", filename, name);
+ return load_image(name, filename);
+ }
return 0;
}
#ifndef _v3_ctrl_h
#define _v3_ctrl_h
+
+/* Global Control IOCTLs */
#define V3_CREATE_GUEST 12
#define V3_FREE_GUEST 13
+#define V3_ADD_MEMORY 50
+#define V3_ADD_PCI_HW_DEV 55
+#define V3_ADD_PCI_USER_DEV 56
+
+/* VM Specific IOCTLs */
/* VM Specific ioctls */
#define V3_VM_CONSOLE_CONNECT 20
#define V3_VM_INSPECT 30
#define V3_VM_DEBUG 31
-#define V3_ADD_MEMORY 50
#define V3_VM_MOVE_CORE 33
char url[256]; /* This might need to be bigger... */
} __attribute__((packed));
+
+
+struct v3_hw_pci_dev {
+ char url[128];
+ unsigned int bus;
+ unsigned int dev;
+ unsigned int func;
+} __attribute__((packed));
+
+
#endif
--- /dev/null
+/* Host PCI User space tool
+ * (c) Jack Lange, 2012
+ * jacklange@cs.pitt.edu
+ */
+#include <stdio.h>
+#include <stdlib.h>
+#include <fcntl.h>
+#include <sys/ioctl.h>
+#include <sys/stat.h>
+#include <sys/types.h>
+#include <unistd.h>
+#include <string.h>
+
+
+#include "v3_ctrl.h"
+
+
+int main(int argc, char ** argv) {
+ int v3_fd = 0;
+ struct v3_hw_pci_dev dev_info;
+ unsigned int bus = 0;
+ unsigned int dev = 0;
+ unsigned int func = 0;
+ int ret = 0;
+
+ if (argc < 3) {
+ printf("Usage: ./v3_pci <name> <bus> <dev> <func>\n");
+ return -1;
+ }
+
+ bus = atoi(argv[2]);
+ dev = atoi(argv[3]);
+ func = atoi(argv[4]);
+
+ strncpy(dev_info.url, argv[1], 128);
+ dev_info.bus = bus;
+ dev_info.dev = dev;
+ dev_info.func = func;
+
+
+ v3_fd = open("/dev/v3vee", O_RDONLY);
+
+ if (v3_fd == -1) {
+ printf("Error opening V3Vee device file\n");
+ return -1;
+ }
+
+
+ ret = ioctl(v3_fd, V3_ADD_PCI_HW_DEV, &dev_info);
+
+
+ if (ret < 0) {
+ printf("Error registering PCI device\n");
+ return -1;
+ }
+
+ close(v3_fd);
+}
--- /dev/null
+/*
+ * This file is part of the Palacios Virtual Machine Monitor developed
+ * by the V3VEE Project with funding from the United States National
+ * Science Foundation and the Department of Energy.
+ *
+ * The V3VEE Project is a joint project between Northwestern University
+ * and the University of New Mexico. You can find out more at
+ * http://www.v3vee.org
+ *
+ * Copyright (c) 2012, Jack Lange <jacklange@cs.pitt.edu>
+ * Copyright (c) 2012, The V3VEE Project <http://www.v3vee.org>
+ * All rights reserved.
+ *
+ * Author: Jack Lange <jacklange@cs.pitt.edu>
+ *
+ * This is free software. You are permitted to use,
+ * redistribute, and modify it as specified in the file "V3VEE_LICENSE".
+ */
+
+
+#include <palacios/vmm.h>
+#include <palacios/vmm_types.h>
+
+
+struct v3_vm_info;
+
+typedef enum { PT_BAR_NONE,
+ PT_BAR_IO,
+ PT_BAR_MEM32,
+ PT_BAR_MEM24,
+ PT_BAR_MEM64_LO,
+ PT_BAR_MEM64_HI,
+ PT_EXP_ROM } pt_bar_type_t;
+
+
+typedef enum { HOST_PCI_CMD_DMA_DISABLE = 1,
+ HOST_PCI_CMD_DMA_ENABLE = 2,
+ HOST_PCI_CMD_INTX_DISABLE = 3,
+ HOST_PCI_CMD_INTX_ENABLE = 4,
+ HOST_PCI_CMD_MSI_DISABLE = 5,
+ HOST_PCI_CMD_MSI_ENABLE = 6,
+ HOST_PCI_CMD_MSIX_DISABLE = 7,
+ HOST_PCI_CMD_MSIX_ENABLE = 8 } host_pci_cmd_t;
+
+struct v3_host_pci_bar {
+ uint32_t size;
+ pt_bar_type_t type;
+
+ /* We store 64 bit memory bar addresses in the high BAR
+ * because they are the last to be updated
+ * This means that the addr field must be 64 bits
+ */
+ uint64_t addr;
+
+ union {
+ uint32_t flags;
+ struct {
+ uint32_t prefetchable : 1;
+ uint32_t cacheable : 1;
+ uint32_t exp_rom_enabled : 1;
+ uint32_t rsvd : 29;
+ } __attribute__((packed));
+ } __attribute__((packed));
+
+
+};
+
+
+
+struct v3_host_pci_dev {
+ struct v3_host_pci_bar bars[6];
+ struct v3_host_pci_bar exp_rom;
+
+ uint8_t cfg_space[256];
+
+ enum {IOMMU, SYMBIOTIC, EMULATED} iface;
+
+ int (*irq_handler)(void * guest_data, uint32_t vec_index);
+
+ void * host_data;
+ void * guest_data;
+};
+
+// For now we just support the single contiguous region
+// This can be updated in the future to support non-contiguous guests
+struct v3_guest_mem_region {
+ uint64_t start;
+ uint64_t end;
+};
+
+
+#ifdef __V3VEE__
+
+#include <devices/pci.h>
+
+
+struct v3_host_pci_dev * v3_host_pci_get_dev(struct v3_vm_info * vm, char * url, void * priv_data);
+
+
+int v3_host_pci_config_write(struct v3_host_pci_dev * v3_dev, uint32_t reg_num, void * src, uint32_t length);
+int v3_host_pci_config_read(struct v3_host_pci_dev * v3_dev, uint32_t reg_num, void * dst, uint32_t length);
+
+int v3_host_pci_cmd_update(struct v3_host_pci_dev * v3_dev, pci_cmd_t cmd, uint64_t arg);
+
+int v3_host_pci_ack_irq(struct v3_host_pci_dev * v3_dev, uint32_t vector);
+
+
+#endif
+
+
+struct v3_host_pci_hooks {
+ struct v3_host_pci_dev * (*request_device)(char * url, void * v3_ctx);
+
+ // emulated interface
+
+ int (*config_write)(struct v3_host_pci_dev * v3_dev, uint32_t reg_num, void * src, uint32_t length);
+ int (*config_read)(struct v3_host_pci_dev * v3_dev, uint32_t reg_num, void * dst, uint32_t length);
+
+ int (*pci_cmd)(struct v3_host_pci_dev * v3_dev, host_pci_cmd_t cmd, uint64_t arg);
+
+ int (*ack_irq)(struct v3_host_pci_dev * v3_dev, uint32_t vector);
+
+
+};
+
+
+
+void V3_Init_Host_PCI(struct v3_host_pci_hooks * hooks);
+
+int V3_get_guest_mem_region(struct v3_vm_info * vm, struct v3_guest_mem_region * region);
+int V3_host_pci_raise_irq(struct v3_host_pci_dev * v3_dev, uint32_t vec_index);
+
--- /dev/null
+/*
+ * This file is part of the Palacios Virtual Machine Monitor developed
+ * by the V3VEE Project with funding from the United States National
+ * Science Foundation and the Department of Energy.
+ *
+ * The V3VEE Project is a joint project between Northwestern University
+ * and the University of New Mexico. You can find out more at
+ * http://www.v3vee.org
+ *
+ * Copyright (c) 2012, Jack Lange <jarusl@cs.northwestern.edu>
+ * Copyright (c) 2012, The V3VEE Project <http://www.v3vee.org>
+ * All rights reserved.
+ *
+ * Author: Jack Lange <jarusl@cs.northwestern.edu>
+ *
+ * This is free software. You are permitted to use,
+ * redistribute, and modify it as specified in the file "V3VEE_LICENSE".
+ */
+
+
+
+#ifndef __SVM_EXITS_H__
+#define __SVM_EXITS_H__
+
+
+#ifdef __V3VEE__
+
+int v3_init_svm_exits(struct v3_vm_info * vm);
+
+
+
+#endif
+
+#endif
#include <palacios/vmm_extensions.h>
#include <palacios/vmm_barrier.h>
#include <palacios/vmm_timeout.h>
-
+#include <palacios/vmm_exits.h>
+#include <palacios/vmm_events.h>
#ifdef V3_CONFIG_TELEMETRY
#include <palacios/vmm_telemetry.h>
struct v3_io_map io_map;
struct v3_msr_map msr_map;
struct v3_cpuid_map cpuid_map;
+ struct v3_exit_map exit_map;
+ struct v3_event_map event_map;
v3_hypercall_map_t hcall_map;
#ifdef __V3VEE__
#include <palacios/vmm_types.h>
+#include <palacios/vmm_lock.h>
struct v3_bitmap {
+ v3_lock_t lock;
int num_bits; // number of valid bit positions in the bitmap
uint8_t * bits; // actual bitmap. Dynamically allocated... ugly
};
--- /dev/null
+/*
+ * This file is part of the Palacios Virtual Machine Monitor developed
+ * by the V3VEE Project with funding from the United States National
+ * Science Foundation and the Department of Energy.
+ *
+ * The V3VEE Project is a joint project between Northwestern University
+ * and the University of New Mexico. You can find out more at
+ * http://www.v3vee.org
+ *
+ * Copyright (c) 2012, Jack Lange <jarusl@cs.northwestern.edu>
+ * Copyright (c) 2012, The V3VEE Project <http://www.v3vee.org>
+ * All rights reserved.
+ *
+ * Author: Jack Lange <jarusl@cs.northwestern.edu>
+ *
+ * This is free software. You are permitted to use,
+ * redistribute, and modify it as specified in the file "V3VEE_LICENSE".
+ */
+
+#ifndef __VMM_EVENTS_H__
+#define __VMM_EVENTS_H__
+
+#ifdef __V3VEE__
+
+#include <palacios/vmm_types.h>
+#include <palacios/vmm_list.h>
+
+struct guest_info;
+struct v3_vm_info;
+
+typedef enum {
+ V3_EVENT_INVALID /* This entry must always be last */
+} v3_event_type_t;
+
+
+struct v3_event_map {
+ struct list_head * events; // array of events
+
+};
+
+
+int v3_init_events(struct v3_vm_info * vm);
+int v3_deinit_events(struct v3_vm_info * vm);
+
+
+struct v3_notifier {
+
+ v3_event_type_t event_type;
+
+ void (*notify)(struct guest_info * core,
+ v3_event_type_t event_type,
+ void * priv_data,
+ void * event_data);
+ void * priv_data;
+
+ struct list_head node;
+
+};
+
+
+struct v3_notifier * v3_subscribe_event(struct v3_vm_info * vm,
+ v3_event_type_t event_type,
+ void (*notify)(struct guest_info * core,
+ v3_event_type_t event_type,
+ void * priv_data,
+ void * event_data),
+ void * priv_data,
+ struct guest_info * current_core);
+
+int v3_unsubscribe_event(struct v3_vm_info * vm, struct v3_notifier * notifier,
+ struct guest_info * current_core);
+
+
+
+#include <palacios/vm_guest.h>
+
+#ifdef __VM_EVENTS_H2___ /* Just ignore the man behind the curtain.... */
+
+static void inline v3_dispatch_event(struct guest_info * core,
+ v3_event_type_t event_type,
+ void * event_data) {
+ struct v3_notifier * tmp_notifier = NULL;
+
+ if (event_type >= V3_EVENT_INVALID) {
+ PrintError("Tried to dispatch illegal event (%d)\n", event_type);
+ return;
+ }
+
+ list_for_each_entry(tmp_notifier, &(core->vm_info->event_map.events[event_type]), node) {
+ tmp_notifier->notify(core, event_type, tmp_notifier->priv_data, event_data);
+ }
+
+}
+
+#endif
+#define __VM_EVENTS_H2___
+
+
+
+
+
+#endif
+
+#endif
--- /dev/null
+/*
+ * This file is part of the Palacios Virtual Machine Monitor developed
+ * by the V3VEE Project with funding from the United States National
+ * Science Foundation and the Department of Energy.
+ *
+ * The V3VEE Project is a joint project between Northwestern University
+ * and the University of New Mexico. You can find out more at
+ * http://www.v3vee.org
+ *
+ * Copyright (c) 2012, Jack Lange <jarusl@cs.northwestern.edu>
+ * Copyright (c) 2012, The V3VEE Project <http://www.v3vee.org>
+ * All rights reserved.
+ *
+ * Author: Jack Lange <jarusl@cs.northwestern.edu>
+ *
+ * This is free software. You are permitted to use,
+ * redistribute, and modify it as specified in the file "V3VEE_LICENSE".
+ */
+
+#ifndef __VMM_EXITS_H__
+#define __VMM_EXITS_H__
+
+
+#ifdef __V3VEE__
+
+#include <palacios/vmm_types.h>
+
+
+
+struct guest_info;
+struct v3_vm_info;
+
+typedef enum { V3_EXIT_RDTSC,
+ V3_EXIT_RDTSCP,
+ V3_EXIT_SWINTR,
+ V3_EXIT_INVALID } v3_exit_type_t;
+
+
+
+
+struct v3_exit_hook {
+
+ int (*enable)(struct guest_info * core, v3_exit_type_t exit_type);
+ int (*disable)(struct guest_info * core, v3_exit_type_t exit_type);
+
+ struct {
+ union {
+ uint32_t flags;
+ struct {
+ uint32_t hooked : 1;
+ uint32_t registered : 1;
+ uint32_t rsvd : 30;
+ } __attribute__((packed));
+ } __attribute__((packed));
+
+ } __attribute__((packed));
+
+
+
+ int (*handler)(struct guest_info * core, v3_exit_type_t exit_type,
+ void * priv_data, void * exit_data);
+ void * priv_data;
+
+};
+
+
+struct v3_exit_map {
+
+ struct v3_exit_hook * exits;
+};
+
+
+
+
+int v3_init_exit_hooks(struct v3_vm_info * vm);
+int v3_deinit_exit_hooks(struct v3_vm_info * vm);
+
+int v3_init_exit_hooks_core(struct guest_info * core);
+
+
+
+int v3_register_exit(struct v3_vm_info * vm, v3_exit_type_t exit_type,
+ int (*enable)(struct guest_info * core, v3_exit_type_t exit_type),
+ int (*disable)(struct guest_info * core, v3_exit_type_t exit_type));
+
+
+
+int v3_dispatch_exit_hook(struct guest_info * core, v3_exit_type_t exit_type, void * exit_data);
+
+
+
+
+
+
+int v3_hook_exit(struct v3_vm_info * vm, v3_exit_type_t exit_type,
+ int (*handler)(struct guest_info * core, v3_exit_type_t exit_type,
+ void * priv_data, void * exit_data),
+ void * priv_data,
+ struct guest_info * current_core);
+
+int v3_unhook_exit(struct v3_vm_info * vm, v3_exit_type_t exit_type, struct guest_info * current_core);
+
+
+#endif
+
+#endif
struct v3_queue * v3_create_queue();
void v3_init_queue(struct v3_queue * queue);
+void v3_deinit_queue(struct v3_queue * queue);
void v3_enqueue(struct v3_queue * queue, addr_t entry);
addr_t v3_dequeue(struct v3_queue * queue);
--- /dev/null
+/*
+ * This file is part of the Palacios Virtual Machine Monitor developed
+ * by the V3VEE Project with funding from the United States National
+ * Science Foundation and the Department of Energy.
+ *
+ * The V3VEE Project is a joint project between Northwestern University
+ * and the University of New Mexico. You can find out more at
+ * http://www.v3vee.org
+ *
+ * Copyright (c) 2012, Jack Lange <jarusl@cs.northwestern.edu>
+ * Copyright (c) 2012, The V3VEE Project <http://www.v3vee.org>
+ * All rights reserved.
+ *
+ * Author: Jack Lange <jarusl@cs.northwestern.edu>
+ *
+ * This is free software. You are permitted to use,
+ * redistribute, and modify it as specified in the file "V3VEE_LICENSE".
+ */
+
+
+
+#ifndef __VMX_EXITS_H__
+#define __VMX_EXITS_H__
+
+
+#ifdef __V3VEE__
+
+int v3_init_vmx_exits(struct v3_vm_info * vm);
+
+
+
+#endif
+
+#endif
state->master_isr &= ~(0x01 << i);
break;
}
- }
+ }
PrintDebug("8259 PIC: Post ISR = %x (wr_Master1)\n", state->master_isr);
} else if (!(cw2->EOI) && (cw2->R) && (cw2->SL)) {
PrintDebug("8259 PIC: Ignoring set-priority, priorities not implemented (level=%d, wr_Master1)\n", cw2->level);
return -1;
}
+ if (cw2->EOI) {
+ if (pic_get_intr_number(core, state) != -1) {
+ PrintError("Interrupt pending after EOI\n");
+ }
+ }
+
+
state->master_ocw2 = cw;
} else if (IS_OCW3(cw)) {
PrintDebug("8259 PIC: Handling OCW3 = %x (wr_Master1)\n", cw);
return -1;
}
+ if (cw2->EOI) {
+ if (pic_get_intr_number(core, state) != -1) {
+ PrintError("Interrupt pending after EOI\n");
+ }
+ }
+
+
+
state->slave_ocw2 = cw;
} else if (IS_OCW3(cw)) {
// Basically sets the IRR/ISR read flag
obj-$(V3_CONFIG_NVRAM) += nvram.o
obj-$(V3_CONFIG_OS_DEBUG) += os_debug.o
obj-$(V3_CONFIG_PCI) += pci.o
+obj-$(V3_CONFIG_HOST_PCI) += host_pci.o
obj-$(V3_CONFIG_PIIX3) += piix3.o
obj-$(V3_CONFIG_SWAPBYPASS_DISK_CACHE) += swapbypass_cache.o
obj-$(V3_CONFIG_SWAPBYPASS_DISK_CACHE2) += swapbypass_cache2.o
obj-$(V3_CONFIG_VGA) += vga.o
-obj-$(V3_CONFIG_PCI_FRONT) += pci_front.o
obj-$(V3_CONFIG_VNET_GUEST_IFACE) += vnet_guest_iface.o
--- /dev/null
+/*
+ * This file is part of the Palacios Virtual Machine Monitor developed
+ * by the V3VEE Project with funding from the United States National
+ * Science Foundation and the Department of Energy.
+ *
+ * The V3VEE Project is a joint project between Northwestern University
+ * and the University of New Mexico. You can find out more at
+ * http://www.v3vee.org
+ *
+ * Copyright (c) 2012, Jack Lange <jacklange@cs.pitt.edu>
+ * Copyright (c) 2012, The V3VEE Project <http://www.v3vee.org>
+ * All rights reserved.
+ *
+ * Author: Jack Lange <jacklange@cs.pitt.edu>
+ *
+ * This is free software. You are permitted to use,
+ * redistribute, and modify it as specified in the file "V3VEE_LICENSE".
+ */
+
+
+/* This is the generic passthrough PCI virtual device */
+
+/*
+ * The basic idea is that we do not change the hardware PCI configuration
+ * Instead we modify the guest environment to map onto the physical configuration
+ *
+ * The pci subsystem handles most of the configuration space, except for the bar registers.
+ * We handle them here, by either letting them go directly to hardware or remapping through virtual hooks
+ *
+ * Memory Bars are always remapped via the shadow map,
+ * IO Bars are selectively remapped through hooks if the guest changes them
+ */
+
+#include <palacios/vmm.h>
+#include <palacios/vmm_dev_mgr.h>
+#include <palacios/vmm_sprintf.h>
+#include <palacios/vmm_lowlevel.h>
+#include <palacios/vm_guest.h> // must include this to avoid dependency issue
+#include <palacios/vmm_symspy.h>
+
+#include <devices/pci.h>
+#include <devices/pci_types.h>
+#include <interfaces/host_pci.h>
+
+#define PCI_BUS_MAX 7
+#define PCI_DEV_MAX 32
+#define PCI_FN_MAX 7
+
+#define PCI_DEVICE 0x0
+#define PCI_PCI_BRIDGE 0x1
+#define PCI_CARDBUS_BRIDGE 0x2
+
+#define PCI_HDR_SIZE 256
+
+
+
+
+struct host_pci_state {
+ // This holds the description of the host PCI device configuration
+ struct v3_host_pci_dev * host_dev;
+
+
+ struct v3_host_pci_bar virt_bars[6];
+ struct v3_host_pci_bar virt_exp_rom;
+
+ struct vm_device * pci_bus;
+ struct pci_device * pci_dev;
+
+ char name[32];
+};
+
+
+
+/*
+static int pci_exp_rom_init(struct vm_device * dev, struct host_pci_state * state) {
+ struct pci_device * pci_dev = state->pci_dev;
+ struct v3_host_pci_bar * hrom = &(state->host_dev->exp_rom);
+
+
+
+ PrintDebug("Adding 32 bit PCI mem region: start=%p, end=%p\n",
+ (void *)(addr_t)hrom->addr,
+ (void *)(addr_t)(hrom->addr + hrom->size));
+
+ if (hrom->exp_rom_enabled) {
+ // only map shadow memory if the ROM is enabled
+
+ v3_add_shadow_mem(dev->vm, V3_MEM_CORE_ANY,
+ hrom->addr,
+ hrom->addr + hrom->size - 1,
+ hrom->addr);
+
+ // Initially the virtual location matches the physical ones
+ memcpy(&(state->virt_exp_rom), hrom, sizeof(struct v3_host_pci_bar));
+
+
+ PrintDebug("phys exp_rom: addr=%p, size=%u\n",
+ (void *)(addr_t)hrom->addr,
+ hrom->size);
+
+
+ // Update the pci subsystem versions
+ pci_dev->config_header.expansion_rom_address = PCI_EXP_ROM_VAL(hrom->addr, hrom->exp_rom_enabled);
+ }
+
+
+
+ return 0;
+}
+*/
+
+
+static int pt_io_read(struct guest_info * core, uint16_t port, void * dst, uint_t length, void * priv_data) {
+ struct v3_host_pci_bar * pbar = (struct v3_host_pci_bar *)priv_data;
+ int port_offset = port % pbar->size;
+
+ if (length == 1) {
+ *(uint8_t *)dst = v3_inb(pbar->addr + port_offset);
+ } else if (length == 2) {
+ *(uint16_t *)dst = v3_inw(pbar->addr + port_offset);
+ } else if (length == 4) {
+ *(uint32_t *)dst = v3_indw(pbar->addr + port_offset);
+ } else {
+ PrintError("Invalid PCI passthrough IO Redirection size read\n");
+ return -1;
+ }
+
+ return length;
+}
+
+
+static int pt_io_write(struct guest_info * core, uint16_t port, void * src, uint_t length, void * priv_data) {
+ struct v3_host_pci_bar * pbar = (struct v3_host_pci_bar *)priv_data;
+ int port_offset = port % pbar->size;
+
+ if (length == 1) {
+ v3_outb(pbar->addr + port_offset, *(uint8_t *)src);
+ } else if (length == 2) {
+ v3_outw(pbar->addr + port_offset, *(uint16_t *)src);
+ } else if (length == 4) {
+ v3_outdw(pbar->addr + port_offset, *(uint32_t *)src);
+ } else {
+ PrintError("Invalid PCI passthrough IO Redirection size write\n");
+ return -1;
+ }
+
+ return length;
+
+}
+
+
+
+static int pci_bar_init(int bar_num, uint32_t * dst, void * private_data) {
+ struct vm_device * dev = (struct vm_device *)private_data;
+ struct host_pci_state * state = (struct host_pci_state *)dev->private_data;
+ struct v3_host_pci_bar * hbar = &(state->host_dev->bars[bar_num]);
+ uint32_t bar_val = 0;
+
+ if (hbar->type == PT_BAR_IO) {
+ int i = 0;
+
+ bar_val = PCI_IO_BAR_VAL(hbar->addr);
+
+ for (i = 0; i < hbar->size; i++) {
+ v3_hook_io_port(dev->vm, hbar->addr + i, NULL, NULL, NULL);
+ }
+ } else if (hbar->type == PT_BAR_MEM32) {
+ bar_val = PCI_MEM32_BAR_VAL(hbar->addr, hbar->prefetchable);
+
+ v3_add_shadow_mem(dev->vm, V3_MEM_CORE_ANY,
+ hbar->addr, hbar->addr + hbar->size - 1,
+ hbar->addr);
+
+ } else if (hbar->type == PT_BAR_MEM24) {
+ bar_val = PCI_MEM24_BAR_VAL(hbar->addr, hbar->prefetchable);
+
+ v3_add_shadow_mem(dev->vm, V3_MEM_CORE_ANY,
+ hbar->addr, hbar->addr + hbar->size - 1,
+ hbar->addr);
+ } else if (hbar->type == PT_BAR_MEM64_LO) {
+ struct v3_host_pci_bar * hi_hbar = &(state->host_dev->bars[bar_num + 1]);
+ bar_val = PCI_MEM64_LO_BAR_VAL(hi_hbar->addr, hbar->prefetchable);
+ } else if (hbar->type == PT_BAR_MEM64_HI) {
+ bar_val = PCI_MEM64_HI_BAR_VAL(hbar->addr, hbar->prefetchable);
+
+ v3_add_shadow_mem(dev->vm, V3_MEM_CORE_ANY,
+ hbar->addr, hbar->addr + hbar->size - 1,
+ hbar->addr);
+ }
+
+
+ memcpy(&(state->virt_bars[bar_num]), hbar, sizeof(struct v3_host_pci_bar));
+
+ *dst = bar_val;
+
+ return 0;
+}
+
+
+
+static int pci_bar_write(int bar_num, uint32_t * src, void * private_data) {
+ struct vm_device * dev = (struct vm_device *)private_data;
+ struct host_pci_state * state = (struct host_pci_state *)dev->private_data;
+
+ struct v3_host_pci_bar * hbar = &(state->host_dev->bars[bar_num]);
+ struct v3_host_pci_bar * vbar = &(state->virt_bars[bar_num]);
+
+
+
+ if (vbar->type == PT_BAR_NONE) {
+ return 0;
+ } else if (vbar->type == PT_BAR_IO) {
+ int i = 0;
+
+ // unhook old ports
+ for (i = 0; i < vbar->size; i++) {
+ if (v3_unhook_io_port(dev->vm, vbar->addr + i) == -1) {
+ PrintError("Could not unhook previously hooked port.... %d (0x%x)\n",
+ (uint32_t)vbar->addr + i, (uint32_t)vbar->addr + i);
+ return -1;
+ }
+ }
+
+ // clear the low bits to match the size
+ vbar->addr = *src & ~(hbar->size - 1);
+
+ // udpate source version
+ *src = PCI_IO_BAR_VAL(vbar->addr);
+
+ PrintDebug("Rehooking passthrough IO ports starting at %d (0x%x)\n",
+ (uint32_t)vbar->addr, (uint32_t)vbar->addr);
+
+ if (vbar->addr == hbar->addr) {
+ // Map the io ports as passthrough
+ for (i = 0; i < hbar->size; i++) {
+ v3_hook_io_port(dev->vm, hbar->addr + i, NULL, NULL, NULL);
+ }
+ } else {
+ // We have to manually handle the io redirection
+ for (i = 0; i < vbar->size; i++) {
+ v3_hook_io_port(dev->vm, vbar->addr + i, pt_io_read, pt_io_write, hbar);
+ }
+ }
+ } else if (vbar->type == PT_BAR_MEM32) {
+ // remove old mapping
+ struct v3_mem_region * old_reg = v3_get_mem_region(dev->vm, V3_MEM_CORE_ANY, vbar->addr);
+
+ if (old_reg == NULL) {
+ // uh oh...
+ PrintError("Could not find PCI Passthrough memory redirection region (addr=0x%x)\n", (uint32_t)vbar->addr);
+ return -1;
+ }
+
+ v3_delete_mem_region(dev->vm, old_reg);
+
+ // clear the low bits to match the size
+ vbar->addr = *src & ~(hbar->size - 1);
+
+ // Set reserved bits
+ *src = PCI_MEM32_BAR_VAL(vbar->addr, hbar->prefetchable);
+
+ PrintDebug("Adding pci Passthrough remapping: start=0x%x, size=%d, end=0x%x (hpa=%p)\n",
+ (uint32_t)vbar->addr, vbar->size, (uint32_t)vbar->addr + vbar->size, (void *)hbar->addr);
+
+ v3_add_shadow_mem(dev->vm, V3_MEM_CORE_ANY,
+ vbar->addr,
+ vbar->addr + vbar->size - 1,
+ hbar->addr);
+
+ } else if (vbar->type == PT_BAR_MEM64_LO) {
+ // We only store the written values here, the actual reconfig comes when the high BAR is updated
+
+ vbar->addr = *src & ~(hbar->size - 1);
+
+ *src = PCI_MEM64_LO_BAR_VAL(vbar->addr, hbar->prefetchable);
+
+
+ } else if (vbar->type == PT_BAR_MEM64_HI) {
+ struct v3_host_pci_bar * lo_vbar = &(state->virt_bars[bar_num - 1]);
+ struct v3_mem_region * old_reg = v3_get_mem_region(dev->vm, V3_MEM_CORE_ANY, vbar->addr);
+
+ if (old_reg == NULL) {
+ // uh oh...
+ PrintError("Could not find PCI Passthrough memory redirection region (addr=%p)\n",
+ (void *)(addr_t)vbar->addr);
+ return -1;
+ }
+
+ // remove old mapping
+ v3_delete_mem_region(dev->vm, old_reg);
+
+ vbar->addr = (((uint64_t)*src) << 32) + lo_vbar->addr;
+
+ // We don't set size, because we assume region is less than 4GB
+ // src does not change, because there are no reserved bits
+
+
+ PrintDebug("Adding pci Passthrough remapping: start=%p, size=%p, end=%p\n",
+ (void *)(addr_t)vbar->addr, (void *)(addr_t)vbar->size,
+ (void *)(addr_t)(vbar->addr + vbar->size));
+
+ if (v3_add_shadow_mem(dev->vm, V3_MEM_CORE_ANY, vbar->addr,
+ vbar->addr + vbar->size - 1, hbar->addr) == -1) {
+
+ PrintDebug("Fail to insert shadow region (%p, %p) -> %p\n",
+ (void *)(addr_t)vbar->addr,
+ (void *)(addr_t)(vbar->addr + vbar->size - 1),
+ (void *)(addr_t)hbar->addr);
+ return -1;
+ }
+
+ } else {
+ PrintError("Unhandled Pasthrough PCI Bar type %d\n", vbar->type);
+ return -1;
+ }
+
+
+ return 0;
+}
+
+
+static int pt_config_write(struct pci_device * pci_dev, uint32_t reg_num, void * src, uint_t length, void * private_data) {
+ struct vm_device * dev = (struct vm_device *)private_data;
+ struct host_pci_state * state = (struct host_pci_state *)dev->private_data;
+
+// V3_Print("Writing host PCI config space update\n");
+
+ // We will mask all operations to the config header itself,
+ // and only allow direct access to the device specific config space
+ if (reg_num < 64) {
+ return 0;
+ }
+
+ return v3_host_pci_config_write(state->host_dev, reg_num, src, length);
+}
+
+
+
+static int pt_config_read(struct pci_device * pci_dev, uint32_t reg_num, void * dst, uint_t length, void * private_data) {
+ struct vm_device * dev = (struct vm_device *)private_data;
+ struct host_pci_state * state = (struct host_pci_state *)dev->private_data;
+
+ // V3_Print("Reading host PCI config space update\n");
+
+ return v3_host_pci_config_read(state->host_dev, reg_num, dst, length);
+}
+
+
+
+
+/* This is really iffy....
+ * It was totally broken before, but it's _not_ totally fixed now
+ * The Expansion rom can be enabled/disabled via software using the low order bit
+ * We should probably handle that somehow here...
+ */
+static int pt_exp_rom_write(struct pci_device * pci_dev, uint32_t * src, void * priv_data) {
+ struct vm_device * dev = (struct vm_device *)(priv_data);
+ struct host_pci_state * state = (struct host_pci_state *)dev->private_data;
+
+ struct v3_host_pci_bar * hrom = &(state->host_dev->exp_rom);
+ struct v3_host_pci_bar * vrom = &(state->virt_exp_rom);
+
+ PrintDebug("exp_rom update: src=0x%x\n", *src);
+ PrintDebug("vrom is size=%u, addr=0x%x\n", vrom->size, (uint32_t)vrom->addr);
+ PrintDebug("hrom is size=%u, addr=0x%x\n", hrom->size, (uint32_t)hrom->addr);
+
+ if (hrom->exp_rom_enabled) {
+ // only remove old mapping if present, I.E. if the rom was enabled previously
+ if (vrom->exp_rom_enabled) {
+ struct v3_mem_region * old_reg = v3_get_mem_region(dev->vm, V3_MEM_CORE_ANY, vrom->addr);
+
+ if (old_reg == NULL) {
+ // uh oh...
+ PrintError("Could not find PCI Passthrough exp_rom_base redirection region (addr=0x%x)\n", (uint32_t)vrom->addr);
+ return -1;
+ }
+
+ v3_delete_mem_region(dev->vm, old_reg);
+ }
+
+
+ vrom->addr = *src & ~(hrom->size - 1);
+
+ // Set flags in actual register value
+ *src = PCI_EXP_ROM_VAL(vrom->addr, (*src & 0x00000001));
+
+ PrintDebug("Cooked src=0x%x\n", *src);
+
+
+ PrintDebug("Adding pci Passthrough exp_rom_base remapping: start=0x%x, size=%u, end=0x%x\n",
+ (uint32_t)vrom->addr, vrom->size, (uint32_t)vrom->addr + vrom->size);
+
+ if (v3_add_shadow_mem(dev->vm, V3_MEM_CORE_ANY, vrom->addr,
+ vrom->addr + vrom->size - 1, hrom->addr) == -1) {
+ PrintError("Failed to remap pci exp_rom: start=0x%x, size=%u, end=0x%x\n",
+ (uint32_t)vrom->addr, vrom->size, (uint32_t)vrom->addr + vrom->size);
+ return -1;
+ }
+ }
+
+ return 0;
+}
+
+
+static int pt_cmd_update(struct pci_device * pci, pci_cmd_t cmd, uint64_t arg, void * priv_data) {
+ struct vm_device * dev = (struct vm_device *)(priv_data);
+ struct host_pci_state * state = (struct host_pci_state *)dev->private_data;
+
+ V3_Print("Host PCI Device: CMD update (%d)(arg=%llu)\n", cmd, arg);
+
+ v3_host_pci_cmd_update(state->host_dev, cmd, arg);
+
+ return 0;
+}
+
+
+static int setup_virt_pci_dev(struct v3_vm_info * vm_info, struct vm_device * dev) {
+ struct host_pci_state * state = (struct host_pci_state *)dev->private_data;
+ struct pci_device * pci_dev = NULL;
+ struct v3_pci_bar bars[6];
+ int bus_num = 0;
+ int i;
+
+ for (i = 0; i < 6; i++) {
+ bars[i].type = PCI_BAR_PASSTHROUGH;
+ bars[i].private_data = dev;
+ bars[i].bar_init = pci_bar_init;
+ bars[i].bar_write = pci_bar_write;
+ }
+
+ pci_dev = v3_pci_register_device(state->pci_bus,
+ PCI_STD_DEVICE,
+ bus_num, -1, 0,
+ state->name, bars,
+ pt_config_write,
+ pt_config_read,
+ pt_cmd_update,
+ pt_exp_rom_write,
+ dev);
+
+
+ state->pci_dev = pci_dev;
+
+ // pci_exp_rom_init(dev, state);
+ pci_dev->config_header.expansion_rom_address = 0;
+
+ v3_pci_enable_capability(pci_dev, PCI_CAP_MSI);
+// v3_pci_enable_capability(pci_dev, PCI_CAP_MSIX);
+ v3_pci_enable_capability(pci_dev, PCI_CAP_PCIE);
+ v3_pci_enable_capability(pci_dev, PCI_CAP_PM);
+
+
+
+ if (state->host_dev->iface == SYMBIOTIC) {
+#ifdef V3_CONFIG_SYMBIOTIC
+ v3_sym_map_pci_passthrough(vm_info, pci_dev->bus_num, pci_dev->dev_num, pci_dev->fn_num);
+#else
+ PrintError("ERROR Symbiotic Passthrough is not enabled\n");
+ return -1;
+#endif
+ }
+
+ return 0;
+}
+
+
+static struct v3_device_ops dev_ops = {
+ .free = NULL,
+};
+
+
+static int irq_ack(struct guest_info * core, uint32_t irq, void * private_data) {
+ struct host_pci_state * state = (struct host_pci_state *)private_data;
+
+
+ // V3_Print("Acking IRQ %d\n", irq);
+ v3_host_pci_ack_irq(state->host_dev, irq);
+
+ return 0;
+}
+
+
+static int irq_handler(void * private_data, uint32_t vec_index) {
+ struct host_pci_state * state = (struct host_pci_state *)private_data;
+ struct v3_irq vec;
+
+ vec.irq = vec_index;
+ vec.ack = irq_ack;
+ vec.private_data = state;
+
+
+ // V3_Print("Raising host PCI IRQ %d\n", vec_index);
+
+ if (state->pci_dev->irq_type == IRQ_NONE) {
+ return 0;
+ } else if (state->pci_dev->irq_type == IRQ_INTX) {
+ v3_pci_raise_acked_irq(state->pci_bus, state->pci_dev, vec);
+ } else {
+ v3_pci_raise_irq(state->pci_bus, state->pci_dev, vec_index);
+ }
+
+ return 0;
+}
+
+
+static int host_pci_init(struct v3_vm_info * vm, v3_cfg_tree_t * cfg) {
+ struct host_pci_state * state = V3_Malloc(sizeof(struct host_pci_state));
+ struct vm_device * dev = NULL;
+ struct vm_device * pci = v3_find_dev(vm, v3_cfg_val(cfg, "bus"));
+ char * dev_id = v3_cfg_val(cfg, "ID");
+ char * url = v3_cfg_val(cfg, "url");
+
+ memset(state, 0, sizeof(struct host_pci_state));
+
+ if (!pci) {
+ PrintError("PCI bus not specified in config file\n");
+ return -1;
+ }
+
+ state->pci_bus = pci;
+ strncpy(state->name, dev_id, 32);
+
+
+ dev = v3_add_device(vm, dev_id, &dev_ops, state);
+
+ if (dev == NULL) {
+ PrintError("Could not attach device %s\n", dev_id);
+ V3_Free(state);
+ return -1;
+ }
+
+ state->host_dev = v3_host_pci_get_dev(vm, url, state);
+
+ if (state->host_dev == NULL) {
+ PrintError("Could not connect to host pci device (%s)\n", url);
+ return -1;
+ }
+
+
+ state->host_dev->irq_handler = irq_handler;
+
+ if (setup_virt_pci_dev(vm, dev) == -1) {
+ PrintError("Could not setup virtual host PCI device\n");
+ return -1;
+ }
+
+ return 0;
+}
+
+
+
+
+device_register("HOST_PCI", host_pci_init)
help
Provides the inspection extension
+config EXT_VMWARE
+ bool "VMWare Persona"
+ default n
+ help
+ Provides a VMWare persona to allow TSC calibration
+
endmenu
obj-$(V3_CONFIG_EXT_VTIME) += ext_vtime.o
obj-$(V3_CONFIG_EXT_INSPECTOR) += ext_inspector.o
obj-$(V3_CONFIG_EXT_MACH_CHECK) += ext_mcheck.o
-
+obj-$(V3_CONFIG_EXT_VMWARE) += ext_vmware.o
--- /dev/null
+
+/*
+ * This file is part of the Palacios Virtual Machine Monitor developed
+ * by the V3VEE Project with funding from the United States National
+ * Science Foundation and the Department of Energy.
+ *
+ * The V3VEE Project is a joint project between Northwestern University
+ * and the University of New Mexico. You can find out more at
+ * http://www.v3vee.org
+ *
+ * Copyright (c) 2012, Jack Lange <jarusl@cs.pitt.edu>
+ * Copyright (c) 2012, The V3VEE Project <http://www.v3vee.org>
+ * All rights reserved.
+ *
+ * Author: Jack Lange <jarusl@cs.pitt.edu>
+ *
+ * This is free software. You are permitted to use,
+ * redistribute, and modify it as specified in the file "V3VEE_LICENSE".
+ */
+
+#include <palacios/vmm.h>
+#include <palacios/vmm_extensions.h>
+#include <palacios/vmm_io.h>
+#include <palacios/vmm_cpuid.h>
+#include <palacios/vm_guest.h>
+
+
+#define VMWARE_CPUID_LEAF 0x40000000
+#define VMWARE_MAGIC 0x564D5868
+#define VMWARE_IO_PORT 0x5658
+
+#define VMWARE_IO_VERSION 10
+#define VMWARE_IO_GETHZ 45
+
+
+static int io_read(struct guest_info * core, uint16_t port, void * dst, uint_t length, void * priv_data) {
+ uint64_t cpu_hz = V3_CPU_KHZ() * 1000;
+ uint32_t magic = (uint32_t)(core->vm_regs.rax);
+ uint32_t cmd = (uint32_t)(core->vm_regs.rcx);
+
+ PrintError("VMWARE IO READ of size %d (command=%d)\n", length, cmd);
+
+
+ if (magic != VMWARE_MAGIC) {
+ PrintError("Invalid VMWARE MAgic number in Persona interface, ignoring for now\n");
+ return length;
+ }
+
+ if (cmd == VMWARE_IO_GETHZ) {
+ // EAX Takes low bytes
+ // EBX takes high bytes
+ core->vm_regs.rax = cpu_hz & 0x00000000ffffffffLL;
+ core->vm_regs.rbx = (cpu_hz >> 32) & 0x00000000ffffffffLL;
+ } else {
+ PrintError("Unhandled VMWARE IO operation\n");
+ return -1;
+ }
+
+ return length;
+}
+
+
+static int io_write(struct guest_info * core, uint16_t port, void * src, uint_t length, void * priv_data) {
+
+ PrintError("VMWARE IO PORT WRITE\n");
+ return -1;
+}
+
+
+static int vmware_cpuid_handler(struct guest_info * core, uint32_t cpuid,
+ uint32_t * eax, uint32_t * ebx,
+ uint32_t * ecx, uint32_t * edx,
+ void * priv_data) {
+
+ // Don't Care (?)
+ *eax = 0;
+
+ // Set VMWARE Vendor string in EBX,ECX,EDX
+ memcpy(ebx, "VMwa", 4);
+ memcpy(ecx, "reVM", 4);
+ memcpy(edx, "ware", 4);
+
+ return 0;
+}
+
+
+
+static int vmware_init(struct v3_vm_info * vm, v3_cfg_tree_t * cfg, void ** priv_data) {
+
+ V3_Print("Using VMWARE virtualization persona\n");
+
+ v3_cpuid_add_fields(vm, 0x00000001,
+ 0, 0,
+ 0, 0,
+ 0x80000000, 0x80000000,
+ 0, 0
+ );
+
+
+ v3_hook_io_port(vm, VMWARE_IO_PORT,
+ io_read, io_write,
+ NULL);
+
+ v3_hook_cpuid(vm, VMWARE_CPUID_LEAF,
+ vmware_cpuid_handler, NULL);
+
+
+ // hook io port
+ // set CPUID hypervisor enabled
+ // set VMWare CPUID
+
+
+ return 0;
+
+
+}
+
+
+
+
+static struct v3_extension_impl vmware_impl = {
+ .name = "VMWARE_IFACE",
+ .init = vmware_init,
+ .deinit = NULL,
+ .core_init = NULL,
+ .core_deinit = NULL,
+ .on_entry = NULL,
+ .on_exit = NULL
+};
+
+
+
+register_extension(&vmware_impl);
to support the internal networking features of Palacios.
+config HOST_PCI
+ bool "Host PCI Support"
+ depends on EXPERIMENTAL
+ default y
+ help
+ This enables host support for passthrough PCI devices
+
config HOST_DEVICE
bool "Host device support"
default n
obj-$(V3_CONFIG_KEYED_STREAMS) += vmm_keyed_stream.o
obj-$(V3_CONFIG_HOST_DEVICE) += vmm_host_dev.o
obj-$(V3_CONFIG_HOST_HYPERCALL) += vmm_host_hypercall.o
+obj-$(V3_CONFIG_HOST_PCI) += host_pci.o
obj-y += null.o
--- /dev/null
+/*
+ * This file is part of the Palacios Virtual Machine Monitor developed
+ * by the V3VEE Project with funding from the United States National
+ * Science Foundation and the Department of Energy.
+ *
+ * The V3VEE Project is a joint project between Northwestern University
+ * and the University of New Mexico. You can find out more at
+ * http://www.v3vee.org
+ *
+ * Copyright (c) 2012, Jack Lange <jacklange@cs.pitt.edu>
+ * Copyright (c) 2012, The V3VEE Project <http://www.v3vee.org>
+ * All rights reserved.
+ *
+ * Author: Jack Lange <jacklange@cs.pitt.edu>
+ *
+ * This is free software. You are permitted to use,
+ * redistribute, and modify it as specified in the file "V3VEE_LICENSE".
+ */
+
+
+#include <interfaces/host_pci.h>
+#include <palacios/vmm.h>
+#include <palacios/vm_guest.h>
+#include <palacios/vmm_mem.h>
+
+
+
+static struct v3_host_pci_hooks * pci_hooks = NULL;
+
+
+
+void V3_Init_Host_PCI(struct v3_host_pci_hooks * hooks) {
+ pci_hooks = hooks;
+ V3_Print("V3 host PCI interface intialized\n");
+ return;
+}
+
+
+/* This is ugly and should be abstracted out to a function in the memory manager */
+int V3_get_guest_mem_region(struct v3_vm_info * vm, struct v3_guest_mem_region * region) {
+
+ if (!vm) {
+ PrintError("Tried to get a nenregion from a NULL vm pointer\n");
+ return -1;
+ }
+
+
+ region->start = vm->mem_map.base_region.host_addr;
+ region->end = vm->mem_map.base_region.host_addr + (vm->mem_map.base_region.guest_end - vm->mem_map.base_region.guest_start);
+
+ return 0;
+}
+
+
+struct v3_host_pci_dev * v3_host_pci_get_dev(struct v3_vm_info * vm,
+ char * url, void * priv_data) {
+
+ struct v3_host_pci_dev * host_dev = NULL;
+
+ if ((!pci_hooks) || (!pci_hooks->request_device)) {
+ PrintError("Host PCI Hooks not initialized\n");
+ return NULL;
+ }
+
+ host_dev = pci_hooks->request_device(url, vm);
+
+ if (host_dev == NULL) {
+ PrintError("Could not find host PCI device (%s)\n", url);
+ return NULL;
+ }
+
+ host_dev->guest_data = priv_data;
+
+ return host_dev;
+
+}
+
+
+int v3_host_pci_config_write(struct v3_host_pci_dev * v3_dev,
+ uint32_t reg_num, void * src,
+ uint32_t length) {
+
+ if ((!pci_hooks) || (!pci_hooks->config_write)) {
+ PrintError("Host PCI hooks not initialized\n");
+ return -1;
+ }
+
+ return pci_hooks->config_write(v3_dev, reg_num, src, length);
+}
+
+
+int v3_host_pci_config_read(struct v3_host_pci_dev * v3_dev,
+ uint32_t reg_num, void * dst,
+ uint32_t length) {
+
+ if ((!pci_hooks) || (!pci_hooks->config_read)) {
+ PrintError("Host PCI hooks not initialized\n");
+ return -1;
+ }
+
+ return pci_hooks->config_read(v3_dev, reg_num, dst, length);
+}
+
+int v3_host_pci_ack_irq(struct v3_host_pci_dev * v3_dev, uint32_t vec_index) {
+
+ if ((!pci_hooks) || (!pci_hooks->ack_irq)) {
+ PrintError("Host PCI hooks not initialized\n");
+ return -1;
+ }
+
+ return pci_hooks->ack_irq(v3_dev, vec_index);
+}
+
+
+
+int v3_host_pci_cmd_update(struct v3_host_pci_dev * v3_dev, pci_cmd_t cmd, uint64_t arg ) {
+
+ if ((!pci_hooks) || (!pci_hooks->pci_cmd)) {
+ PrintError("Host PCI hooks not initialized\n");
+ return -1;
+ }
+
+ return pci_hooks->pci_cmd(v3_dev, cmd, arg);
+}
+
+
+
+
+
+int V3_host_pci_raise_irq(struct v3_host_pci_dev * v3_dev, uint32_t vec_index) {
+ if (!v3_dev->irq_handler) {
+ PrintError("No interrupt registerd for host pci device\n");
+ return -1;
+ }
+
+ return v3_dev->irq_handler(v3_dev->guest_data, vec_index);
+}
+
vmm_bitmap.o \
vmm_barrier.o \
vmm_timeout.o \
+ vmm_exits.o \
+ vmm_events.o
obj-$(V3_CONFIG_XED) += vmm_xed.o
svm_pause.o \
svm_wbinvd.o \
svm_handler.o \
+ svm_exits.o \
vmcb.o
obj-$(V3_CONFIG_VMX) += vmx.o \
vmcs.o \
vmx_ctrl_regs.o \
vmx_assist.o \
- vmx_ept.o
+ vmx_ept.o \
+ vmx_exits.o
*/
+
#include <palacios/svm.h>
#include <palacios/vmm.h>
#include <palacios/vmm_barrier.h>
#include <palacios/vmm_debug.h>
+
+
#ifdef V3_CONFIG_CHECKPOINT
#include <palacios/vmm_checkpoint.h>
#endif
v3_svm_config_tsc_virtualization(struct guest_info * info) {
vmcb_ctrl_t * ctrl_area = GET_VMCB_CTRL_AREA((vmcb_t*)(info->vmm_data));
+
if (info->time_state.flags & VM_TIME_TRAP_RDTSC) {
ctrl_area->instrs.RDTSC = 1;
ctrl_area->svm_instrs.RDTSCP = 1;
} else {
ctrl_area->instrs.RDTSC = 0;
ctrl_area->svm_instrs.RDTSCP = 0;
+
if (info->time_state.flags & VM_TIME_TSC_PASSTHROUGH) {
ctrl_area->TSC_OFFSET = 0;
} else {
--- /dev/null
+/*
+ * This file is part of the Palacios Virtual Machine Monitor developed
+ * by the V3VEE Project with funding from the United States National
+ * Science Foundation and the Department of Energy.
+ *
+ * The V3VEE Project is a joint project between Northwestern University
+ * and the University of New Mexico. You can find out more at
+ * http://www.v3vee.org
+ *
+ * Copyright (c) 2012, Jack Lange <jarusl@cs.northwestern.edu>
+ * Copyright (c) 2012, The V3VEE Project <http://www.v3vee.org>
+ * All rights reserved.
+ *
+ * Author: Jack Lange <jarusl@cs.northwestern.edu>
+ *
+ * This is free software. You are permitted to use,
+ * redistribute, and modify it as specified in the file "V3VEE_LICENSE".
+ */
+
+
+#include <palacios/vmm.h>
+#include <palacios/vmcb.h>
+#include <palacios/vmm_exits.h>
+
+
+
+static int enable_exit(struct guest_info * core, v3_exit_type_t exit_type) {
+ vmcb_ctrl_t * ctrl_area = GET_VMCB_CTRL_AREA((vmcb_t *)(core->vmm_data));
+
+ switch (exit_type) {
+
+ case V3_EXIT_RDTSC:
+ ctrl_area->instrs.RDTSC = 1;
+ break;
+ case V3_EXIT_RDTSCP:
+ ctrl_area->svm_instrs.RDTSCP = 1;
+ break;
+
+ default:
+ PrintError("Unhandled Exit Type (%d)\n", exit_type);
+ return -1;
+ }
+
+ return 0;
+}
+
+
+static int disable_exit(struct guest_info * core, v3_exit_type_t exit_type) {
+ vmcb_ctrl_t * ctrl_area = GET_VMCB_CTRL_AREA((vmcb_t *)(core->vmm_data));
+
+ switch (exit_type) {
+
+ case V3_EXIT_RDTSC:
+ ctrl_area->instrs.RDTSC = 0;
+ break;
+ case V3_EXIT_RDTSCP:
+ ctrl_area->svm_instrs.RDTSCP = 0;
+ break;
+
+ default:
+ PrintError("Unhandled Exit Type (%d)\n", exit_type);
+ return -1;
+ }
+
+ return 0;
+
+}
+
+
+int v3_init_svm_exits(struct v3_vm_info * vm) {
+
+ int ret = 0;
+
+ ret |= v3_register_exit(vm, V3_EXIT_RDTSC, enable_exit, disable_exit);
+ ret |= v3_register_exit(vm, V3_EXIT_RDTSCP, enable_exit, disable_exit);
+
+ return ret;
+}
PrintDebug("RDTSCP\n");
#endif
if (v3_handle_rdtscp(info) == -1) {
- PrintError("Error Handling RDTSCP instruction\n");
+ PrintError("Error handling RDTSCP instruction\n");
return -1;
}
+
break;
case SVM_EXIT_SHUTDOWN:
PrintDebug("Guest-initiated shutdown\n");
#include <palacios/svm.h>
#include <palacios/svm_io.h>
#include <palacios/svm_msr.h>
+#include <palacios/svm_exits.h>
#endif
#ifdef V3_CONFIG_VMX
#include <palacios/vmx.h>
#include <palacios/vmx_io.h>
#include <palacios/vmx_msr.h>
+#include <palacios/vmx_exits.h>
#endif
int v3_init_vm(struct v3_vm_info * vm) {
extern v3_cpu_arch_t v3_mach_type;
-
+ v3_init_events(vm);
#ifdef V3_CONFIG_TELEMETRY
v3_init_telemetry(vm);
#endif
+ v3_init_exit_hooks(vm);
v3_init_hypercall_map(vm);
v3_init_io_map(vm);
v3_init_msr_map(vm);
v3_init_cpuid_map(vm);
v3_init_host_events(vm);
+
v3_init_intr_routers(vm);
v3_init_ext_manager(vm);
case V3_SVM_REV3_CPU:
v3_init_svm_io_map(vm);
v3_init_svm_msr_map(vm);
+ v3_init_svm_exits(vm);
break;
#endif
#ifdef V3_CONFIG_VMX
case V3_VMX_EPT_UG_CPU:
v3_init_vmx_io_map(vm);
v3_init_vmx_msr_map(vm);
+ v3_init_vmx_exits(vm);
break;
#endif
default:
v3_deinit_io_map(vm);
v3_deinit_hypercall_map(vm);
+ v3_deinit_exit_hooks(vm);
+
#ifdef V3_CONFIG_TELEMETRY
v3_deinit_telemetry(vm);
#endif
+ v3_deinit_events(vm);
return 0;
PrintError("Invalid CPU Type 0x%x\n", v3_mach_type);
return -1;
}
+
+ v3_init_exit_hooks_core(core);
+
return 0;
}
int v3_stop_vm(struct v3_vm_info * vm) {
+ if ((vm->run_state |= VM_RUNNING) &&
+ (vm->run_state != VM_SIMULATING)) {
+ PrintError("Tried to stop VM in invalid runstate (%d)\n", vm->run_state);
+ return -1;
+ }
+
vm->run_state = VM_STOPPED;
// Sanity check to catch any weird execution states
int i = 0;
// deinitialize guest (free memory, etc...)
+ if ((vm->run_state != VM_STOPPED) &&
+ (vm->run_state != VM_ERROR)) {
+ PrintError("Tried to Free VM in invalid runstate (%d)\n", vm->run_state);
+ return -1;
+ }
+
v3_free_vm_devices(vm);
// free cores
int v3_raise_barrier(struct v3_vm_info * vm_info, struct guest_info * local_core) {
int ret = 0;
+
+ if ((vm_info->run_state != VM_RUNNING) ||
+ (vm_info->run_state != VM_SIMULATING)) {
+ return 0;
+ }
+
ret = v3_raise_barrier_nowait(vm_info, local_core);
if (ret != 0) {
int v3_lower_barrier(struct v3_vm_info * vm_info) {
struct v3_barrier * barrier = &(vm_info->barrier);
+
+ if ((vm_info->run_state != VM_RUNNING) ||
+ (vm_info->run_state != VM_SIMULATING)) {
+ return 0;
+ }
+
// Clear the active flag, so cores won't wait
barrier->active = 0;
int v3_bitmap_init(struct v3_bitmap * bitmap, int num_bits) {
int num_bytes = (num_bits / 8) + ((num_bits % 8) > 0);
+ v3_lock_init(&(bitmap->lock));
bitmap->num_bits = num_bits;
bitmap->bits = V3_Malloc(num_bytes);
+
if (bitmap->bits == NULL) {
PrintError("Could not allocate bitmap of %d bits\n", num_bits);
return -1;
void v3_bitmap_deinit(struct v3_bitmap * bitmap) {
+ v3_lock_deinit(&(bitmap->lock));
V3_Free(bitmap->bits);
}
int major = index / 8;
int minor = index % 8;
int old_val = 0;
+ uint32_t flags = 0;
if (index > (bitmap->num_bits - 1)) {
PrintError("Index out of bitmap range: (pos = %d) (num_bits = %d)\n",
return -1;
}
+
+ flags = v3_lock_irqsave(bitmap->lock);
+
old_val = (bitmap->bits[major] & (0x1 << minor));
bitmap->bits[major] |= (0x1 << minor);
+ v3_unlock_irqrestore(bitmap->lock, flags);
+
return old_val;
}
int major = index / 8;
int minor = index % 8;
int old_val = 0;
+ uint32_t flags = 0;
if (index > (bitmap->num_bits - 1)) {
PrintError("Index out of bitmap range: (pos = %d) (num_bits = %d)\n",
return -1;
}
+ flags = v3_lock_irqsave(bitmap->lock);
+
old_val = (bitmap->bits[major] & (0x1 << minor));
bitmap->bits[major] &= ~(0x1 << minor);
+ v3_unlock_irqrestore(bitmap->lock, flags);
+
return old_val;
}
--- /dev/null
+/*
+ * This file is part of the Palacios Virtual Machine Monitor developed
+ * by the V3VEE Project with funding from the United States National
+ * Science Foundation and the Department of Energy.
+ *
+ * The V3VEE Project is a joint project between Northwestern University
+ * and the University of New Mexico. You can find out more at
+ * http://www.v3vee.org
+ *
+ * Copyright (c) 2012, Jack Lange <jarusl@cs.northwestern.edu>
+ * Copyright (c) 2012, The V3VEE Project <http://www.v3vee.org>
+ * All rights reserved.
+ *
+ * Author: Jack Lange <jarusl@cs.northwestern.edu>
+ *
+ * This is free software. You are permitted to use,
+ * redistribute, and modify it as specified in the file "V3VEE_LICENSE".
+ */
+
+
+
+#include <palacios/vmm_events.h>
+#include <palacios/vm_guest.h>
+
+
+
+int v3_init_events(struct v3_vm_info * vm) {
+ struct v3_event_map * map = &(vm->event_map);
+ int i = 0;
+
+ map->events = V3_Malloc(sizeof(struct list_head) * V3_EVENT_INVALID);
+
+ if (map->events == NULL) {
+ PrintError("Error: could not allocate event map\n");
+ return -1;
+ }
+
+ for (i = 0; i < V3_EVENT_INVALID; i++) {
+ INIT_LIST_HEAD(&(map->events[i]));
+ }
+
+ return 0;
+}
+
+int v3_deinit_events(struct v3_vm_info * vm) {
+ struct v3_event_map * map = &(vm->event_map);
+ int i = 0;
+
+ for (i = 0; i < V3_EVENT_INVALID; i++) {
+ if (!list_empty(&(map->events[i]))) {
+ struct v3_notifier * tmp_notifier = NULL;
+ struct v3_notifier * safe_notifier = NULL;
+ PrintError("Found orphan notifier for event %d. Probable memory leak detected.\n", i);
+
+ list_for_each_entry_safe(tmp_notifier, safe_notifier, &(map->events[i]), node) {
+ list_del(&(tmp_notifier->node));
+ V3_Free(tmp_notifier);
+ }
+ }
+ }
+
+
+ V3_Free(map->events);
+
+ return 0;
+
+}
+
+
+struct v3_notifier * v3_subscribe_event(struct v3_vm_info * vm,
+ v3_event_type_t event_type,
+ void (*notify)(struct guest_info * core,
+ v3_event_type_t event_type,
+ void * priv_data,
+ void * event_data),
+ void * priv_data,
+ struct guest_info * current_core) {
+ struct v3_event_map * map = &(vm->event_map);
+ struct v3_notifier * notifier = NULL;
+
+ if (event_type >= V3_EVENT_INVALID) {
+ PrintError("Tried to request illegal event (%d)\n", event_type);
+ return NULL;
+ }
+
+ notifier = V3_Malloc(sizeof(struct v3_notifier));
+
+ if (notifier == NULL) {
+ PrintError("Error: Could not allocate notifier\n");
+ return NULL;
+ }
+
+ memset(notifier, 0, sizeof(struct v3_notifier));
+
+ notifier->notify = notify;
+ notifier->priv_data = priv_data;
+ notifier->event_type = event_type;
+
+ while (v3_raise_barrier(vm, current_core) == -1);
+ list_add(&(notifier->node), &(map->events[event_type]));
+ v3_lower_barrier(vm);
+
+ return notifier;;
+}
+
+
+int v3_unsubscribe_event(struct v3_vm_info * vm, struct v3_notifier * notifier,
+ struct guest_info * current_core) {
+ struct v3_event_map * map = &(vm->event_map);
+ struct v3_notifier * tmp_notifier = NULL;
+ struct v3_notifier * safe_notifier = NULL;
+
+ if (notifier == NULL) {
+ PrintError("Could not unsubscribe invalid event notifier\n");
+ return -1;
+ }
+
+ if (notifier->event_type >= V3_EVENT_INVALID) {
+ PrintError("Could not unsubscribe from invalid event\n");
+ return -1;
+ }
+
+ while (v3_raise_barrier(vm, current_core) == -1);
+ list_for_each_entry_safe(tmp_notifier, safe_notifier, &(map->events[notifier->event_type]), node) {
+ if (tmp_notifier == notifier) {
+ list_del(&(tmp_notifier->node));
+ V3_Free(tmp_notifier);
+ }
+ }
+ v3_lower_barrier(vm);
+
+ return 0;
+}
--- /dev/null
+/*
+ * This file is part of the Palacios Virtual Machine Monitor developed
+ * by the V3VEE Project with funding from the United States National
+ * Science Foundation and the Department of Energy.
+ *
+ * The V3VEE Project is a joint project between Northwestern University
+ * and the University of New Mexico. You can find out more at
+ * http://www.v3vee.org
+ *
+ * Copyright (c) 2012, Jack Lange <jarusl@cs.northwestern.edu>
+ * Copyright (c) 2012, The V3VEE Project <http://www.v3vee.org>
+ * All rights reserved.
+ *
+ * Author: Jack Lange <jarusl@cs.northwestern.edu>
+ *
+ * This is free software. You are permitted to use,
+ * redistribute, and modify it as specified in the file "V3VEE_LICENSE".
+ */
+
+
+#include <palacios/vmm.h>
+#include <palacios/vmm_exits.h>
+#include <palacios/vm_guest.h>
+
+
+int v3_init_exit_hooks(struct v3_vm_info * vm) {
+ struct v3_exit_map * map = &(vm->exit_map);
+
+ map->exits = V3_Malloc(sizeof(struct v3_exit_hook) * V3_EXIT_INVALID);
+
+ if (map->exits == NULL) {
+ PrintError("Error allocating exit map\n");
+ return -1;
+ }
+
+ memset(map->exits, 0, sizeof(struct v3_exit_hook) * V3_EXIT_INVALID);
+
+
+ return 0;
+}
+
+int v3_deinit_exit_hooks(struct v3_vm_info * vm) {
+ struct v3_exit_map * map = &(vm->exit_map);
+
+ V3_Free(map->exits);
+
+ return 0;
+}
+
+
+
+
+int v3_init_exit_hooks_core(struct guest_info * core) {
+ struct v3_vm_info * vm = core->vm_info;
+ struct v3_exit_map * map = &(vm->exit_map);
+ struct v3_exit_hook * hook = NULL;
+ int i = 0;
+
+ for (i = 0; i < V3_EXIT_INVALID; i++) {
+ hook = &(map->exits[i]);
+
+ if (hook->hooked) {
+ if (hook->enable(core, i) != 0) {
+ PrintError("Error could not enable exit hook %d on core %d\n", i, core->vcpu_id);
+ return -1;
+ }
+ }
+ }
+
+ return 0;
+}
+
+int v3_deinit_exit_hooks_core(struct guest_info * core) {
+
+ return 0;
+}
+
+
+
+int v3_dispatch_exit_hook(struct guest_info * core, v3_exit_type_t exit_type, void * exit_data) {
+ struct v3_exit_map * map = &(core->vm_info->exit_map);
+ struct v3_exit_hook * hook = NULL;
+
+ if (exit_type >= V3_EXIT_INVALID) {
+ PrintError("Error: Tried to dispatch invalid exit type (%d)\n", exit_type);
+ return -1;
+ }
+
+ hook = &(map->exits[exit_type]);
+
+ if (hook->hooked == 0) {
+ PrintError("Tried to dispatch an unhooked exit (%d)\n", exit_type);
+ return -1;
+ }
+
+ return hook->handler(core, exit_type, hook->priv_data, exit_data);
+
+}
+
+
+int v3_register_exit(struct v3_vm_info * vm, v3_exit_type_t exit_type,
+ int (*enable)(struct guest_info * core, v3_exit_type_t exit_type),
+ int (*disable)(struct guest_info * core, v3_exit_type_t exit_type)) {
+ struct v3_exit_map * map = &(vm->exit_map);
+ struct v3_exit_hook * hook = NULL;
+
+ if (exit_type >= V3_EXIT_INVALID) {
+ PrintError("Error: Tried to register invalid exit type (%d)\n", exit_type);
+ return -1;
+ }
+
+ hook = &(map->exits[exit_type]);
+
+ if (hook->registered == 1) {
+ PrintError("Tried to reregister an exit (%d)\n", exit_type);
+ return -1;
+ }
+
+ hook->registered = 1;
+ hook->enable = enable;
+ hook->disable = disable;
+
+
+ return 0;
+}
+
+
+int v3_hook_exit(struct v3_vm_info * vm, v3_exit_type_t exit_type,
+ int (*handler)(struct guest_info * core, v3_exit_type_t exit_type,
+ void * priv_data, void * exit_data),
+ void * priv_data,
+ struct guest_info * current_core) {
+ struct v3_exit_map * map = &(vm->exit_map);
+ struct v3_exit_hook * hook = NULL;
+
+
+ if (exit_type >= V3_EXIT_INVALID) {
+ PrintError("Error: Tried to hook invalid exit type (%d)\n", exit_type);
+ return -1;
+ }
+
+ hook = &(map->exits[exit_type]);
+
+ if (hook->registered == 0) {
+ PrintError("Tried to hook unregistered exit (%d)\n", exit_type);
+ return -1;
+ }
+
+ if (hook->hooked != 0) {
+ PrintError("Tried to rehook exit (%d)\n", exit_type);
+ return -1;
+ }
+
+
+ hook->hooked = 1;
+ hook->handler = handler;
+ hook->priv_data = priv_data;
+
+ if (vm->run_state != VM_INVALID) {
+ int i = 0;
+
+ while (v3_raise_barrier(vm, current_core) == -1);
+
+ for (i = 0; i < vm->num_cores; i++) {
+
+ if (hook->enable(&(vm->cores[i]), exit_type) != 0) {
+ PrintError("Error could not enable exit hook %d on core %d\n", exit_type, i);
+ v3_lower_barrier(vm);
+ return -1;
+ }
+ }
+
+ v3_lower_barrier(vm);
+ }
+
+ return 0;
+}
+
+
+
+int v3_unhook_exit(struct v3_vm_info * vm, v3_exit_type_t exit_type, struct guest_info * current_core) {
+ struct v3_exit_map * map = &(vm->exit_map);
+ struct v3_exit_hook * hook = NULL;
+
+
+ if (exit_type >= V3_EXIT_INVALID) {
+ PrintError("Error: Tried to unhook invalid exit type (%d)\n", exit_type);
+ return -1;
+ }
+
+ hook = &(map->exits[exit_type]);
+
+ if (hook->registered == 0) {
+ PrintError("Tried to unhook an unregistered exit (%d)\n", exit_type);
+ return -1;
+ }
+
+ if (hook->hooked == 0) {
+ PrintError("Tried to unhook and unhooked exit (%d)\n", exit_type);
+ return -1;
+ }
+
+
+ hook->hooked = 0;
+ hook->handler = NULL;
+ hook->priv_data = NULL;
+
+
+ if (vm->run_state != VM_INVALID) {
+ int i = 0;
+
+ while (v3_raise_barrier(vm, current_core) == -1);
+
+ for (i = 0; i < vm->num_cores; i++) {
+
+ if (hook->disable(&(vm->cores[i]), exit_type) != 0) {
+ PrintError("Error could not enable exit hook %d on core %d\n", exit_type, i);
+ v3_lower_barrier(vm);
+ return -1;
+ }
+ }
+
+ v3_lower_barrier(vm);
+ }
+
+ return 0;
+}
*entry = 0;
return PT_ENTRY_NOT_PRESENT;
} else if (pdpe_entry->large_page) {
- PrintError("1 Gigabyte pages not supported\n");
- V3_ASSERT(0);
- return -1;
+ pdpe64_1GB_t * large_pdp = (pdpe64_1GB_t *)pdpe_entry;
+
+ *entry = BASE_TO_PAGE_ADDR_1GB(large_pdp->page_base_addr);
+
+ return PT_ENTRY_LARGE_PAGE;
} else {
*entry = BASE_TO_PAGE_ADDR(pdpe_entry->pd_base_addr);
return PT_ENTRY_PAGE;
#include <palacios/vmm_queue.h>
+
+
+
void v3_init_queue(struct v3_queue * queue) {
queue->num_entries = 0;
INIT_LIST_HEAD(&(queue->entries));
return tmp_queue;
}
+void v3_deinit_queue(struct v3_queue * queue) {
+ while (v3_dequeue(queue)) {
+ PrintError("ERROR: Freeing non-empty queue. PROBABLE MEMORY LEAK DETECTED\n");
+ }
+
+ v3_lock_deinit(&(queue->lock));
+}
+
+
+
+
void v3_enqueue(struct v3_queue * queue, addr_t entry) {
struct v3_queue_entry * q_entry = V3_Malloc(sizeof(struct v3_queue_entry));
unsigned int flags = 0;
--- /dev/null
+/*
+ * This file is part of the Palacios Virtual Machine Monitor developed
+ * by the V3VEE Project with funding from the United States National
+ * Science Foundation and the Department of Energy.
+ *
+ * The V3VEE Project is a joint project between Northwestern University
+ * and the University of New Mexico. You can find out more at
+ * http://www.v3vee.org
+ *
+ * Copyright (c) 2012, Jack Lange <jarusl@cs.northwestern.edu>
+ * Copyright (c) 2012, The V3VEE Project <http://www.v3vee.org>
+ * All rights reserved.
+ *
+ * Author: Jack Lange <jarusl@cs.northwestern.edu>
+ *
+ * This is free software. You are permitted to use,
+ * redistribute, and modify it as specified in the file "V3VEE_LICENSE".
+ */
+
+
+#include <palacios/vmm.h>
+#include <palacios/vmcb.h>
+#include <palacios/vmm_exits.h>
+
+
+
+static int enable_exit(struct guest_info * core, v3_exit_type_t exit_type) {
+ vmcb_ctrl_t * ctrl_area = GET_VMCB_CTRL_AREA((vmcb_t *)(core->vmm_data));
+
+ switch (exit_type) {
+
+ case V3_EXIT_RDTSC:
+ ctrl_area->instrs.RDTSC = 1;
+ break;
+ case V3_EXIT_RDTSCP:
+ ctrl_area->svm_instrs.RDTSCP = 1;
+ break;
+
+ default:
+ PrintError("Unhandled Exit Type (%d)\n", exit_type);
+ return -1;
+ }
+
+ return 0;
+}
+
+
+static int disable_exit(struct guest_info * core, v3_exit_type_t exit_type) {
+ vmcb_ctrl_t * ctrl_area = GET_VMCB_CTRL_AREA((vmcb_t *)(core->vmm_data));
+
+ switch (exit_type) {
+
+ case V3_EXIT_RDTSC:
+ ctrl_area->instrs.RDTSC = 0;
+ break;
+ case V3_EXIT_RDTSCP:
+ ctrl_area->svm_instrs.RDTSCP = 0;
+ break;
+
+ default:
+ PrintError("Unhandled Exit Type (%d)\n", exit_type);
+ return -1;
+ }
+
+ return 0;
+
+}
+
+
+int v3_init_vmx_exits(struct v3_vm_info * vm) {
+
+ int ret = 0;
+
+ ret |= v3_register_exit(vm, V3_EXIT_RDTSC, enable_exit, disable_exit);
+ ret |= v3_register_exit(vm, V3_EXIT_RDTSCP, enable_exit, disable_exit);
+
+ return ret;
+}
}
-void v3_deinit_vnet(){
+void v3_deinit_vnet() {
+
+ v3_deinit_queue(vnet_state.poll_devs);
+ Vnet_Free(vnet_state.poll_devs);
PrintDebug("Stopping flush thread\n");
// This will pause until the flush thread is gone