int xc_dom_alloc_segment(struct xc_dom_image *dom,
struct xc_dom_seg *seg, char *name,
xen_vaddr_t start, xen_vaddr_t size)
{
unsigned int page_size = XC_DOM_PAGE_SIZE(dom);
xen_pfn_t pages;
void *ptr;
if ( start && xc_dom_alloc_pad(dom, start) )
return -1;
pages = (size + page_size - 1) / page_size;
start = dom->virt_alloc_end;
seg->pfn = dom->pfn_alloc_end;
seg->pages = pages;
if ( xc_dom_chk_alloc_pages(dom, name, pages) )
return -1;
/* map and clear pages */
ptr = xc_dom_seg_to_ptr(dom, seg);
if ( ptr == NULL )
return -1;
memset(ptr, 0, pages * page_size);
seg->vstart = start;
seg->vend = dom->virt_alloc_end;
DOMPRINTF("%-20s: %-12s : 0x%" PRIx64 " -> 0x%" PRIx64
" (pfn 0x%" PRIpfn " + 0x%" PRIpfn " pages)",
__FUNCTION__, name, seg->vstart, seg->vend, seg->pfn, pages);
return 0;
}
int xc_dom_alloc_segment(struct xc_dom_image *dom,
struct xc_dom_seg *seg, char *name,
xen_vaddr_t start, xen_vaddr_t size)
{
unsigned int page_size = XC_DOM_PAGE_SIZE(dom);
xen_pfn_t pages = (size + page_size - 1) / page_size;
xen_pfn_t pfn;
void *ptr;
if ( start == 0 )
start = dom->virt_alloc_end;
if ( start & (page_size - 1) )
{
xc_dom_panic(dom->xch, XC_INTERNAL_ERROR,
"%s: segment start isn't page aligned (0x%" PRIx64 ")",
__FUNCTION__, start);
return -1;
}
if ( start < dom->virt_alloc_end )
{
xc_dom_panic(dom->xch, XC_INTERNAL_ERROR,
"%s: segment start too low (0x%" PRIx64 " < 0x%" PRIx64
")", __FUNCTION__, start, dom->virt_alloc_end);
return -1;
}
seg->vstart = start;
pfn = (seg->vstart - dom->parms.virt_base) / page_size;
seg->pfn = pfn + dom->rambase_pfn;
if ( pages > dom->total_pages || /* multiple test avoids overflow probs */
pfn > dom->total_pages ||
pages > dom->total_pages - pfn)
{
xc_dom_panic(dom->xch, XC_OUT_OF_MEMORY,
"%s: segment %s too large (0x%"PRIpfn" > "
"0x%"PRIpfn" - 0x%"PRIpfn" pages)",
__FUNCTION__, name, pages, dom->total_pages, pfn);
return -1;
}
seg->vend = start + pages * page_size;
dom->virt_alloc_end = seg->vend;
if (dom->allocate)
dom->allocate(dom, dom->virt_alloc_end);
DOMPRINTF("%-20s: %-12s : 0x%" PRIx64 " -> 0x%" PRIx64
" (pfn 0x%" PRIpfn " + 0x%" PRIpfn " pages)",
__FUNCTION__, name, seg->vstart, seg->vend, seg->pfn, pages);
/* map and clear pages */
ptr = xc_dom_seg_to_ptr(dom, seg);
if ( ptr == NULL )
return -1;
memset(ptr, 0, pages * page_size);
return 0;
}
static int xc_dom_load_elf_kernel(struct xc_dom_image *dom)
{
struct elf_binary *elf = dom->private_loader;
elf->dest = xc_dom_seg_to_ptr(dom, &dom->kernel_seg);
elf_load_binary(elf);
if ( dom->parms.bsd_symtab )
xc_dom_load_elf_symtab(dom, elf, 1);
return 0;
}
static int alloc_magic_pages(struct xc_dom_image *dom)
{
struct ft_cxt devtree;
void *guest_devtree;
unsigned long shadow_mb;
int rma_pages;
int rc;
/* Allocate special pages from the end of the RMA. */
rma_pages = 1 << (dom->realmodearea_log - PAGE_SHIFT);
dom->shared_info_pfn = --rma_pages;
dom->console_pfn = --rma_pages;
dom->xenstore_pfn = --rma_pages;
/* Gather shadow allocation info for the device tree. */
rc = xc_shadow_control(dom->guest_xc, dom->guest_domid,
XEN_DOMCTL_SHADOW_OP_GET_ALLOCATION, NULL, 0,
&shadow_mb, 0, NULL);
if (rc < 0 || shadow_mb == 0) {
xc_dom_printf("Couldn't get shadow allocation size or it was 0.\n");
return rc;
}
/* Build device tree. */
rc = make_devtree(&devtree, dom, shadow_mb);
if (rc < 0) {
xc_dom_printf("Failed to create flattened device tree.\n");
return rc;
}
/* Find a spot for it. */
rc = xc_dom_alloc_segment(dom, &dom->devicetree_seg, "devtree", 0,
devtree.bph->totalsize);
if (rc)
goto out;
/* Copy the device tree into place. */
guest_devtree = xc_dom_seg_to_ptr(dom, &dom->devicetree_seg);
if (!guest_devtree) {
xc_dom_printf("Couldn't map guest memory for device tree.\n");
rc = -1;
goto out;
}
memcpy(guest_devtree, devtree.bph, devtree.bph->totalsize);
out:
free_devtree(&devtree);
return rc;
}
int xc_dom_alloc_segment(struct xc_dom_image *dom,
struct xc_dom_seg *seg, char *name,
xen_vaddr_t start, xen_vaddr_t size)
{
unsigned int page_size = XC_DOM_PAGE_SIZE(dom);
xen_pfn_t pages = (size + page_size - 1) / page_size;
void *ptr;
if ( start == 0 )
start = dom->virt_alloc_end;
if ( start & (page_size - 1) )
{
xc_dom_panic(XC_INTERNAL_ERROR,
"%s: segment start isn't page aligned (0x%" PRIx64 ")\n",
__FUNCTION__, start);
return -1;
}
if ( start < dom->virt_alloc_end )
{
xc_dom_panic(XC_INTERNAL_ERROR,
"%s: segment start too low (0x%" PRIx64 " < 0x%" PRIx64
")\n", __FUNCTION__, start, dom->virt_alloc_end);
return -1;
}
seg->vstart = start;
seg->vend = start + pages * page_size;
seg->pfn = (seg->vstart - dom->parms.virt_base) / page_size;
dom->virt_alloc_end = seg->vend;
if (dom->allocate)
dom->allocate(dom, dom->virt_alloc_end);
xc_dom_printf("%-20s: %-12s : 0x%" PRIx64 " -> 0x%" PRIx64
" (pfn 0x%" PRIpfn " + 0x%" PRIpfn " pages)\n",
__FUNCTION__, name, seg->vstart, seg->vend, seg->pfn, pages);
/* map and clear pages */
ptr = xc_dom_seg_to_ptr(dom, seg);
if ( ptr == NULL )
return -1;
memset(ptr, 0, pages * page_size);
return 0;
}
static int xc_dom_load_zimage_kernel(struct xc_dom_image *dom)
{
void *dst;
DOMPRINTF_CALLED(dom->xch);
dst = xc_dom_seg_to_ptr(dom, &dom->kernel_seg);
if ( dst == NULL )
{
DOMPRINTF("%s: xc_dom_seg_to_ptr(dom, &dom->kernel_seg) => NULL",
__func__);
return -1;
}
DOMPRINTF("%s: kernel sed %#"PRIx64"-%#"PRIx64,
__func__, dom->kernel_seg.vstart, dom->kernel_seg.vend);
DOMPRINTF("%s: copy %zd bytes from blob %p to dst %p",
__func__, dom->kernel_size, dom->kernel_blob, dst);
memcpy(dst, dom->kernel_blob, dom->kernel_size);
return 0;
}
int xc_dom_build_image(struct xc_dom_image *dom)
{
unsigned int page_size;
xc_dom_printf("%s: called\n", __FUNCTION__);
/* check for arch hooks */
if ( dom->arch_hooks == NULL )
{
xc_dom_panic(XC_INTERNAL_ERROR, "%s: arch hooks not set\n",
__FUNCTION__);
goto err;
}
page_size = XC_DOM_PAGE_SIZE(dom);
/* load kernel */
if ( xc_dom_alloc_segment(dom, &dom->kernel_seg, "kernel",
dom->kernel_seg.vstart,
dom->kernel_seg.vend -
dom->kernel_seg.vstart) != 0 )
goto err;
if ( dom->kernel_loader->loader(dom) != 0 )
goto err;
/* load ramdisk */
if ( dom->ramdisk_blob )
{
size_t unziplen, ramdisklen;
void *ramdiskmap;
unziplen = xc_dom_check_gzip(dom->ramdisk_blob, dom->ramdisk_size);
ramdisklen = unziplen ? unziplen : dom->ramdisk_size;
if ( xc_dom_alloc_segment(dom, &dom->ramdisk_seg, "ramdisk", 0,
ramdisklen) != 0 )
goto err;
ramdiskmap = xc_dom_seg_to_ptr(dom, &dom->ramdisk_seg);
if ( unziplen )
{
if ( xc_dom_do_gunzip(dom->ramdisk_blob, dom->ramdisk_size,
ramdiskmap, ramdisklen) == -1 )
goto err;
}
else
memcpy(ramdiskmap, dom->ramdisk_blob, dom->ramdisk_size);
}
/* allocate other pages */
if ( dom->arch_hooks->alloc_magic_pages(dom) != 0 )
goto err;
if ( dom->arch_hooks->count_pgtables )
{
dom->arch_hooks->count_pgtables(dom);
if ( (dom->pgtables > 0) &&
(xc_dom_alloc_segment(dom, &dom->pgtables_seg, "page tables", 0,
dom->pgtables * page_size) != 0) )
goto err;
}
if ( dom->alloc_bootstack )
dom->bootstack_pfn = xc_dom_alloc_page(dom, "boot stack");
xc_dom_printf("%-20s: virt_alloc_end : 0x%" PRIx64 "\n",
__FUNCTION__, dom->virt_alloc_end);
xc_dom_printf("%-20s: virt_pgtab_end : 0x%" PRIx64 "\n",
__FUNCTION__, dom->virt_pgtab_end);
return 0;
err:
return -1;
}
void kexec(void *kernel, long kernel_size, void *module, long module_size, char *cmdline, unsigned long flags)
{
struct xc_dom_image *dom;
int rc;
domid_t domid = DOMID_SELF;
xen_pfn_t pfn;
xc_interface *xc_handle;
unsigned long i;
void *seg;
xen_pfn_t boot_page_mfn = virt_to_mfn(&_boot_page);
char features[] = "";
struct mmu_update *m2p_updates;
unsigned long nr_m2p_updates;
DEBUG("booting with cmdline %s\n", cmdline);
xc_handle = xc_interface_open(0,0,0);
dom = xc_dom_allocate(xc_handle, cmdline, features);
dom->allocate = kexec_allocate;
/* We are using guest owned memory, therefore no limits. */
xc_dom_kernel_max_size(dom, 0);
xc_dom_ramdisk_max_size(dom, 0);
dom->kernel_blob = kernel;
dom->kernel_size = kernel_size;
dom->ramdisk_blob = module;
dom->ramdisk_size = module_size;
dom->flags = flags;
dom->console_evtchn = start_info.console.domU.evtchn;
dom->xenstore_evtchn = start_info.store_evtchn;
tpm_hash2pcr(dom, cmdline);
if ( (rc = xc_dom_boot_xen_init(dom, xc_handle, domid)) != 0 ) {
grub_printf("xc_dom_boot_xen_init returned %d\n", rc);
errnum = ERR_BOOT_FAILURE;
goto out;
}
if ( (rc = xc_dom_parse_image(dom)) != 0 ) {
grub_printf("xc_dom_parse_image returned %d\n", rc);
errnum = ERR_BOOT_FAILURE;
goto out;
}
#ifdef __i386__
if (strcmp(dom->guest_type, "xen-3.0-x86_32p")) {
grub_printf("can only boot x86 32 PAE kernels, not %s\n", dom->guest_type);
errnum = ERR_EXEC_FORMAT;
goto out;
}
#endif
#ifdef __x86_64__
if (strcmp(dom->guest_type, "xen-3.0-x86_64")) {
grub_printf("can only boot x86 64 kernels, not %s\n", dom->guest_type);
errnum = ERR_EXEC_FORMAT;
goto out;
}
#endif
/* equivalent of xc_dom_mem_init */
dom->arch_hooks = xc_dom_find_arch_hooks(xc_handle, dom->guest_type);
dom->total_pages = start_info.nr_pages;
/* equivalent of arch_setup_meminit */
/* setup initial p2m */
dom->p2m_host = malloc(sizeof(*dom->p2m_host) * dom->total_pages);
/* Start with our current P2M */
for (i = 0; i < dom->total_pages; i++)
dom->p2m_host[i] = pfn_to_mfn(i);
if ( (rc = xc_dom_build_image(dom)) != 0 ) {
grub_printf("xc_dom_build_image returned %d\n", rc);
errnum = ERR_BOOT_FAILURE;
goto out;
}
/* copy hypercall page */
/* TODO: domctl instead, but requires privileges */
if (dom->parms.virt_hypercall != -1) {
pfn = PHYS_PFN(dom->parms.virt_hypercall - dom->parms.virt_base);
memcpy((void *) pages[pfn], hypercall_page, PAGE_SIZE);
}
/* Equivalent of xc_dom_boot_image */
dom->shared_info_mfn = PHYS_PFN(start_info.shared_info);
if (!xc_dom_compat_check(dom)) {
grub_printf("xc_dom_compat_check failed\n");
errnum = ERR_EXEC_FORMAT;
goto out;
}
/* Move current console, xenstore and boot MFNs to the allocated place */
do_exchange(dom, dom->console_pfn, start_info.console.domU.mfn);
do_exchange(dom, dom->xenstore_pfn, start_info.store_mfn);
DEBUG("virt base at %llx\n", dom->parms.virt_base);
DEBUG("bootstack_pfn %lx\n", dom->bootstack_pfn);
_boot_target = dom->parms.virt_base + PFN_PHYS(dom->bootstack_pfn);
DEBUG("_boot_target %lx\n", _boot_target);
do_exchange(dom, PHYS_PFN(_boot_target - dom->parms.virt_base),
virt_to_mfn(&_boot_page));
/* Make sure the bootstrap page table does not RW-map any of our current
* page table frames */
kexec_allocate(dom, dom->virt_pgtab_end);
if ( (rc = xc_dom_update_guest_p2m(dom))) {
grub_printf("xc_dom_update_guest_p2m returned %d\n", rc);
errnum = ERR_BOOT_FAILURE;
goto out;
}
if ( dom->arch_hooks->setup_pgtables )
if ( (rc = dom->arch_hooks->setup_pgtables(dom))) {
grub_printf("setup_pgtables returned %d\n", rc);
errnum = ERR_BOOT_FAILURE;
goto out;
}
/* start info page */
#undef start_info
if ( dom->arch_hooks->start_info )
dom->arch_hooks->start_info(dom);
#define start_info (start_info_union.start_info)
xc_dom_log_memory_footprint(dom);
/* Unmap libxc's projection of the boot page table */
seg = xc_dom_seg_to_ptr(dom, &dom->pgtables_seg);
munmap(seg, dom->pgtables_seg.vend - dom->pgtables_seg.vstart);
/* Unmap day0 pages to avoid having a r/w mapping of the future page table */
for (pfn = 0; pfn < allocated; pfn++)
munmap((void*) pages[pfn], PAGE_SIZE);
/* Pin the boot page table base */
if ( (rc = pin_table(dom->xch,
#ifdef __i386__
MMUEXT_PIN_L3_TABLE,
#endif
#ifdef __x86_64__
MMUEXT_PIN_L4_TABLE,
#endif
xc_dom_p2m_host(dom, dom->pgtables_seg.pfn),
dom->guest_domid)) != 0 ) {
grub_printf("pin_table(%lx) returned %d\n", xc_dom_p2m_host(dom,
dom->pgtables_seg.pfn), rc);
errnum = ERR_BOOT_FAILURE;
goto out_remap;
}
/* We populate the Mini-OS page table here so that boot.S can just call
* update_va_mapping to project itself there. */
need_pgt(_boot_target);
DEBUG("day0 pages %lx\n", allocated);
DEBUG("boot target page %lx\n", _boot_target);
DEBUG("boot page %p\n", &_boot_page);
DEBUG("boot page mfn %lx\n", boot_page_mfn);
_boot_page_entry = PFN_PHYS(boot_page_mfn) | L1_PROT;
DEBUG("boot page entry %llx\n", _boot_page_entry);
_boot_oldpdmfn = virt_to_mfn(start_info.pt_base);
DEBUG("boot old pd mfn %lx\n", _boot_oldpdmfn);
DEBUG("boot pd virt %lx\n", dom->pgtables_seg.vstart);
_boot_pdmfn = dom->p2m_host[PHYS_PFN(dom->pgtables_seg.vstart - dom->parms.virt_base)];
DEBUG("boot pd mfn %lx\n", _boot_pdmfn);
_boot_stack = _boot_target + PAGE_SIZE;
DEBUG("boot stack %lx\n", _boot_stack);
_boot_start_info = dom->parms.virt_base + PFN_PHYS(dom->start_info_pfn);
DEBUG("boot start info %lx\n", _boot_start_info);
_boot_start = dom->parms.virt_entry;
DEBUG("boot start %lx\n", _boot_start);
/* Keep only useful entries */
for (nr_m2p_updates = pfn = 0; pfn < start_info.nr_pages; pfn++)
if (dom->p2m_host[pfn] != pfn_to_mfn(pfn))
nr_m2p_updates++;
m2p_updates = malloc(sizeof(*m2p_updates) * nr_m2p_updates);
for (i = pfn = 0; pfn < start_info.nr_pages; pfn++)
if (dom->p2m_host[pfn] != pfn_to_mfn(pfn)) {
m2p_updates[i].ptr = PFN_PHYS(dom->p2m_host[pfn]) | MMU_MACHPHYS_UPDATE;
m2p_updates[i].val = pfn;
i++;
}
for (i = 0; i < blk_nb; i++)
shutdown_blkfront(blk_dev[i]);
if (net_dev)
shutdown_netfront(net_dev);
if (kbd_dev)
shutdown_kbdfront(kbd_dev);
stop_kernel();
/* Update M2P */
if ((rc = HYPERVISOR_mmu_update(m2p_updates, nr_m2p_updates, NULL, DOMID_SELF)) < 0) {
xprintk("Could not update M2P\n");
ASSERT(0);
}
xprintk("go!\n");
/* Jump to trampoline boot page */
_boot();
ASSERT(0);
out_remap:
for (pfn = 0; pfn < allocated; pfn++)
do_map_frames(pages[pfn], &pages_mfns[pfn], 1, 0, 0, DOMID_SELF, 0, L1_PROT);
out:
xc_dom_release(dom);
for (pfn = 0; pfn < allocated; pfn++)
free_page((void*)pages[pfn]);
free(pages);
free(pages_mfns);
pages = NULL;
pages_mfns = NULL;
allocated = 0;
xc_interface_close(xc_handle );
}