xc_ia64_linux_save.c

xen虚拟机源代码安装包

/******************************************************************************
 * xc_ia64_linux_save.c
 *
 * Save the state of a running Linux session.
 *
 * Copyright (c) 2003, K A Fraser.
 *  Rewritten for ia64 by Tristan Gingold <[e-mail]>
 *
 * Copyright (c) 2007 Isaku Yamahata <[e-mail]>
 *   Use foreign p2m exposure.
 *   VTi domain support.
 */

#include <inttypes.h>
#include <time.h>
#include <stdlib.h>
#include <unistd.h>
#include <sys/time.h>

#include "xg_private.h"
#include "xc_ia64.h"
#include "xc_ia64_save_restore.h"
#include "xc_efi.h"
#include "xen/hvm/params.h"

/*
** Default values for important tuning parameters. Can override by passing
** non-zero replacement values to xc_linux_save().
**
** XXX SMH: should consider if want to be able to override MAX_MBIT_RATE too.
**
*/
#define DEF_MAX_ITERS    (4 - 1)        /* limit us to 4 times round loop  */
#define DEF_MAX_FACTOR   3              /* never send more than 3x nr_pfns */

/*
** During (live) save/migrate, we maintain a number of bitmaps to track
** which pages we have to send, and to skip.
*/
static inline int test_bit(int nr, volatile void * addr)
{
    return (BITMAP_ENTRY(nr, addr) >> BITMAP_SHIFT(nr)) & 1;
}

static inline void clear_bit(int nr, volatile void * addr)
{
    BITMAP_ENTRY(nr, addr) &= ~(1UL << BITMAP_SHIFT(nr));
}

static inline void set_bit(int nr, volatile void * addr)
{
    BITMAP_ENTRY(nr, addr) |= (1UL << BITMAP_SHIFT(nr));
}

static int
suspend_and_state(int (*suspend)(int), int xc_handle, int io_fd,
                  int dom, xc_dominfo_t *info)
{
    int i = 0;

    if (!(*suspend)(dom)) {
        ERROR("Suspend request failed");
        return -1;
    }

retry:

    if (xc_domain_getinfo(xc_handle, dom, 1, info) != 1) {
        ERROR("Could not get domain info");
        return -1;
    }

    if (info->shutdown && info->shutdown_reason == SHUTDOWN_suspend)
        return 0; // success

    if (info->paused) {
        // try unpausing domain, wait, and retest
        xc_domain_unpause(xc_handle, dom);

        ERROR("Domain was paused. Wait and re-test.");
        usleep(10000);  // 10ms

        goto retry;
    }


    if(++i < 100) {
        ERROR("Retry suspend domain.");
        usleep(10000);  // 10ms
        goto retry;
    }

    ERROR("Unable to suspend domain.");

    return -1;
}

static inline int
md_is_not_ram(const efi_memory_desc_t *md)
{
    return ((md->type != EFI_CONVENTIONAL_MEMORY) ||
            (md->attribute != EFI_MEMORY_WB) ||
            (md->num_pages == 0));
}

/*
 * Send through a list of all the PFNs that were not in map at the close.
 * We send pages which was allocated. However balloon driver may 
 * decreased after sending page. So we have to check the freed
 * page after pausing the domain.
 */
static int
xc_ia64_send_unallocated_list(int xc_handle, int io_fd, 
                              struct xen_ia64_p2m_table *p2m_table,
                              xen_ia64_memmap_info_t *memmap_info, 
                              void *memmap_desc_start, void *memmap_desc_end)
{
    void *p;
    efi_memory_desc_t *md;

    unsigned long N;
    unsigned long pfntab[1024];
    unsigned int j;

    j = 0;
    for (p = memmap_desc_start;
         p < memmap_desc_end;
         p += memmap_info->efi_memdesc_size) {
        md = p;

        if (md_is_not_ram(md))
            continue;

        for (N = md->phys_addr >> PAGE_SHIFT;
             N < (md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT)) >>
                 PAGE_SHIFT;
             N++) {
            if (!xc_ia64_p2m_allocated(p2m_table, N))
                j++;
        }
    }
    if (write_exact(io_fd, &j, sizeof(unsigned int))) {
        ERROR("Error when writing to state file (6a)");
        return -1;
    }
        
    j = 0;
    for (p = memmap_desc_start;
         p < memmap_desc_end;
         p += memmap_info->efi_memdesc_size) {
        md = p;

        if (md_is_not_ram(md))
            continue;

        for (N = md->phys_addr >> PAGE_SHIFT;
             N < (md->phys_addr + (md->num_pages << EFI_PAGE_SHIFT)) >>
                 PAGE_SHIFT;
             N++) {
            if (!xc_ia64_p2m_allocated(p2m_table, N))
                pfntab[j++] = N;
            if (j == sizeof(pfntab)/sizeof(pfntab[0])) {
                if (write_exact(io_fd, &pfntab, sizeof(pfntab[0]) * j)) {
                    ERROR("Error when writing to state file (6b)");
                    return -1;
                }
                j = 0;
            }
        }
    }
    if (j > 0) {
        if (write_exact(io_fd, &pfntab, sizeof(pfntab[0]) * j)) {
            ERROR("Error when writing to state file (6c)");
            return -1;
        }
    }

    return 0;
}

static int
xc_ia64_send_vcpu_context(int xc_handle, int io_fd, uint32_t dom,
                          uint32_t vcpu, vcpu_guest_context_any_t *ctxt_any)
{
    vcpu_guest_context_t *ctxt = &ctxt_any->c;
    if (xc_vcpu_getcontext(xc_handle, dom, vcpu, ctxt_any)) {
        ERROR("Could not get vcpu context");
        return -1;
    }

    if (write_exact(io_fd, ctxt, sizeof(*ctxt))) {
        ERROR("Error when writing to state file (1)");
        return -1;
    }

    fprintf(stderr, "ip=%016lx, b0=%016lx\n", ctxt->regs.ip, ctxt->regs.b[0]);
    return 0;
}

static int
xc_ia64_send_shared_info(int xc_handle, int io_fd, shared_info_t *live_shinfo)
{
    if (write_exact(io_fd, live_shinfo, PAGE_SIZE)) {
        ERROR("Error when writing to state file (1)");
        return -1;
    }
    return 0;
}

static int
xc_ia64_send_vcpumap(int xc_handle, int io_fd, uint32_t dom,
                     const xc_dominfo_t *info, uint64_t max_virt_cpus,
                     uint64_t **vcpumapp)
{
    int rc = -1;
    unsigned int i;
    unsigned long vcpumap_size;
    uint64_t *vcpumap = NULL;

    vcpumap_size = (max_virt_cpus + 1 + sizeof(vcpumap[0]) - 1) /
        sizeof(vcpumap[0]);
    vcpumap = malloc(vcpumap_size);
    if (vcpumap == NULL) {
        ERROR("memory alloc for vcpumap");
        goto out;
    }
    memset(vcpumap, 0, vcpumap_size);

    for (i = 0; i <= info->max_vcpu_id; i++) {
        xc_vcpuinfo_t vinfo;
        if ((xc_vcpu_getinfo(xc_handle, dom, i, &vinfo) == 0) && vinfo.online)
            __set_bit(i, vcpumap);
    }

    if (write_exact(io_fd, &max_virt_cpus, sizeof(max_virt_cpus))) {
        ERROR("write max_virt_cpus");
        goto out;
    }

    if (write_exact(io_fd, vcpumap, vcpumap_size)) {
        ERROR("write vcpumap");
        goto out;
    }

    rc = 0;

 out:
    if (rc != 0 && vcpumap != NULL) {
        free(vcpumap);
        vcpumap = NULL;
    }
    *vcpumapp = vcpumap;
    return rc;
}


static int
xc_ia64_pv_send_context(int xc_handle, int io_fd, uint32_t dom,
                        const xc_dominfo_t *info, shared_info_t *live_shinfo)
{
    int rc = -1;
    unsigned int i;

    /* vcpu map */
    uint64_t *vcpumap = NULL;
    if (xc_ia64_send_vcpumap(xc_handle, io_fd, dom, info, MAX_VIRT_CPUS,
                             &vcpumap))
        goto out;

    /* vcpu context */
    for (i = 0; i <= info->max_vcpu_id; i++) {
        /* A copy of the CPU context of the guest. */
        vcpu_guest_context_any_t ctxt_any;
        vcpu_guest_context_t *ctxt = &ctxt_any.c;

        char *mem;

        if (!__test_bit(i, vcpumap))
            continue;

        if (xc_ia64_send_vcpu_context(xc_handle, io_fd, dom, i, &ctxt_any))
            goto out;

        mem = xc_map_foreign_range(xc_handle, dom, PAGE_SIZE,
                                   PROT_READ|PROT_WRITE, ctxt->privregs_pfn);
        if (mem == NULL) {
            ERROR("cannot map privreg page");
            goto out;
        }
        if (write_exact(io_fd, mem, PAGE_SIZE)) {
            ERROR("Error when writing privreg to state file (5)");
            munmap(mem, PAGE_SIZE);
            goto out;
        }
        munmap(mem, PAGE_SIZE);
    }    

    rc = xc_ia64_send_shared_info(xc_handle, io_fd, live_shinfo);

 out:
    if (vcpumap != NULL)
        free(vcpumap);
    return rc;
}

static int
xc_ia64_hvm_send_context(int xc_handle, int io_fd, uint32_t dom,
                         const xc_dominfo_t *info, shared_info_t *live_shinfo)
{
    int rc = -1;
    unsigned int i;

    /* vcpu map */
    uint64_t *vcpumap = NULL;

    /* HVM: magic frames for ioreqs and xenstore comms */
    const int hvm_params[] = {
        HVM_PARAM_STORE_PFN,
        HVM_PARAM_IOREQ_PFN,
        HVM_PARAM_BUFIOREQ_PFN,
        HVM_PARAM_BUFPIOREQ_PFN,
    };
    const int NR_PARAMS = sizeof(hvm_params) / sizeof(hvm_params[0]);
    /* ioreq_pfn, bufioreq_pfn, store_pfn */
    uint64_t magic_pfns[NR_PARAMS];

    /* HVM: a buffer for holding HVM contxt */
    uint64_t rec_size;
    uint64_t hvm_buf_size = 0;
    uint8_t *hvm_buf = NULL;

    if (xc_ia64_send_shared_info(xc_handle, io_fd, live_shinfo))
        return -1;

    /* vcpu map */
    if (xc_ia64_send_vcpumap(xc_handle, io_fd, dom, info, MAX_VIRT_CPUS,
                             &vcpumap))
        goto out;

    /* vcpu context */
    for (i = 0; i <= info->max_vcpu_id; i++) {
        /* A copy of the CPU context of the guest. */
        vcpu_guest_context_any_t ctxt_any;

        if (!__test_bit(i, vcpumap))
            continue;

        if (xc_ia64_send_vcpu_context(xc_handle, io_fd, dom, i, &ctxt_any))
            goto out;

        /* system context of vcpu is sent as hvm context. */
    }    

    /* Save magic-page locations. */
    memset(magic_pfns, 0, sizeof(magic_pfns));
    for (i = 0; i < NR_PARAMS; i++) {
        if (xc_get_hvm_param(xc_handle, dom, hvm_params[i], &magic_pfns[i])) {
            PERROR("Error when xc_get_hvm_param");
            goto out;
        }
    }

    if (write_exact(io_fd, magic_pfns, sizeof(magic_pfns))) {
        ERROR("Error when writing to state file (7)");
        goto out;
    }

    /* Need another buffer for HVM context */
    hvm_buf_size = xc_domain_hvm_getcontext(xc_handle, dom, 0, 0);
    if (hvm_buf_size == -1) {
        ERROR("Couldn't get HVM context size from Xen");
        goto out;
    }

    hvm_buf = malloc(hvm_buf_size);
    if (!hvm_buf) {
        ERROR("Couldn't allocate memory");
        goto out;
    }

    /* Get HVM context from Xen and save it too */
    rec_size = xc_domain_hvm_getcontext(xc_handle, dom, hvm_buf, hvm_buf_size);
    if (rec_size == -1) {
        ERROR("HVM:Could not get hvm buffer");
        goto out;
    }
        
    if (write_exact(io_fd, &rec_size, sizeof(rec_size))) {
        ERROR("error write hvm buffer size");
        goto out;
    }
        
    if (write_exact(io_fd, hvm_buf, rec_size)) {
        ERROR("write HVM info failed!\n");
        goto out;
    }

    rc = 0;
out:
    if (hvm_buf != NULL)
        free(hvm_buf);
    if (vcpumap != NULL)
        free(vcpumap);
    return rc;
}

int
xc_domain_save(int xc_handle, int io_fd, uint32_t dom, uint32_t max_iters,
               uint32_t max_factor, uint32_t flags, int (*suspend)(int),
               int hvm, void *(*init_qemu_maps)(int, unsigned),
               void (*qemu_flip_buffer)(int, int))
{
    DECLARE_DOMCTL;