3 * Copyright (c) 2004 Christian Limpach.
4 * Copyright (c) 2004-2006,2008 Kip Macy
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. All advertising materials mentioning features or use of this software
16 * must display the following acknowledgement:
17 * This product includes software developed by Christian Limpach.
18 * 4. The name of the author may not be used to endorse or promote products
19 * derived from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
22 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
23 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
24 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
25 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
26 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
27 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
28 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
29 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
30 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
33 #include <sys/cdefs.h>
34 __FBSDID("$FreeBSD$");
36 #include <sys/param.h>
37 #include <sys/systm.h>
41 #include <sys/mount.h>
42 #include <sys/malloc.h>
43 #include <sys/mutex.h>
44 #include <sys/kernel.h>
46 #include <sys/reboot.h>
47 #include <sys/sysproto.h>
49 #include <machine/xen/xen-os.h>
53 #include <machine/segments.h>
54 #include <machine/pcb.h>
55 #include <machine/stdarg.h>
56 #include <machine/vmparam.h>
57 #include <machine/cpu.h>
58 #include <machine/intr_machdep.h>
59 #include <machine/md_var.h>
60 #include <machine/asmacros.h>
64 #include <xen/hypervisor.h>
65 #include <machine/xen/xenvar.h>
66 #include <machine/xen/xenfunc.h>
67 #include <machine/xen/xenpmap.h>
68 #include <machine/xen/xenfunc.h>
69 #include <xen/interface/memory.h>
70 #include <machine/xen/features.h>
72 #include <machine/privatespace.h>
76 #include <vm/vm_page.h>
79 #define IDTVEC(name) __CONCAT(X,name)
82 IDTVEC(div), IDTVEC(dbg), IDTVEC(nmi), IDTVEC(bpt), IDTVEC(ofl),
83 IDTVEC(bnd), IDTVEC(ill), IDTVEC(dna), IDTVEC(fpusegm),
84 IDTVEC(tss), IDTVEC(missing), IDTVEC(stk), IDTVEC(prot),
85 IDTVEC(page), IDTVEC(mchk), IDTVEC(rsvd), IDTVEC(fpu), IDTVEC(align),
86 IDTVEC(xmm), IDTVEC(lcall_syscall), IDTVEC(int0x80_syscall);
90 start_info_t *xen_start_info;
91 shared_info_t *HYPERVISOR_shared_info;
92 xen_pfn_t *xen_machine_phys = machine_to_phys_mapping;
93 xen_pfn_t *xen_phys_machine;
94 xen_pfn_t *xen_pfn_to_mfn_frame_list[16];
95 xen_pfn_t *xen_pfn_to_mfn_frame_list_list;
96 int preemptable, init_first;
97 extern unsigned int avail_space;
106 CTR0(KTR_SPARE2, "ni_cli disabling interrupts");
107 __asm__("pushl %edx;"
120 __asm__("pushl %edx;"
132 * Modify the cmd_line by converting ',' to NULLs so that it is in a format
133 * suitable for the static env vars.
136 xen_setbootenv(char *cmd_line)
140 /* Skip leading spaces */
141 for (; *cmd_line == ' '; cmd_line++);
143 printk("xen_setbootenv(): cmd_line='%s'\n", cmd_line);
145 for (cmd_line_next = cmd_line; strsep(&cmd_line_next, ",") != NULL;);
154 {"boot_askname", RB_ASKNAME},
155 {"boot_single", RB_SINGLE},
156 {"boot_nosync", RB_NOSYNC},
157 {"boot_halt", RB_ASKNAME},
158 {"boot_serial", RB_SERIAL},
159 {"boot_cdrom", RB_CDROM},
160 {"boot_gdb", RB_GDB},
161 {"boot_gdb_pause", RB_RESERVED1},
162 {"boot_verbose", RB_VERBOSE},
163 {"boot_multicons", RB_MULTIPLE},
168 xen_boothowto(char *envp)
172 /* get equivalents from the environment */
173 for (i = 0; howto_names[i].ev != NULL; i++)
174 if (getenv(howto_names[i].ev) != NULL)
175 howto |= howto_names[i].mask;
179 #define PRINTK_BUFSIZE 1024
181 printk(const char *fmt, ...)
185 static char buf[PRINTK_BUFSIZE];
188 retval = vsnprintf(buf, PRINTK_BUFSIZE - 1, fmt, ap);
191 (void)HYPERVISOR_console_write(buf, retval);
195 #define XPQUEUE_SIZE 128
203 /* per-cpu queues and indices */
205 static struct mmu_log xpq_queue_log[MAX_VIRT_CPUS][XPQUEUE_SIZE];
208 static int xpq_idx[MAX_VIRT_CPUS];
209 static mmu_update_t xpq_queue[MAX_VIRT_CPUS][XPQUEUE_SIZE];
211 #define XPQ_QUEUE_LOG xpq_queue_log[vcpu]
212 #define XPQ_QUEUE xpq_queue[vcpu]
213 #define XPQ_IDX xpq_idx[vcpu]
214 #define SET_VCPU() int vcpu = smp_processor_id()
217 static mmu_update_t xpq_queue[XPQUEUE_SIZE];
219 static struct mmu_log xpq_queue_log[XPQUEUE_SIZE];
221 static int xpq_idx = 0;
223 #define XPQ_QUEUE_LOG xpq_queue_log
224 #define XPQ_QUEUE xpq_queue
225 #define XPQ_IDX xpq_idx
229 #define XPQ_IDX_INC atomic_add_int(&XPQ_IDX, 1);
235 int _xpq_idx = XPQ_IDX;
241 printk("xen_dump_queue(): %u entries\n", _xpq_idx);
242 for (i = 0; i < _xpq_idx; i++) {
243 printk(" val: %llx ptr: %llx\n", XPQ_QUEUE[i].val, XPQ_QUEUE[i].ptr);
250 _xen_flush_queue(void)
253 int _xpq_idx = XPQ_IDX;
257 if (__predict_true(gdtset))
258 CRITICAL_ASSERT(curthread);
262 /* Make sure index is cleared first to avoid double updates. */
263 error = HYPERVISOR_mmu_update((mmu_update_t *)&XPQ_QUEUE,
264 _xpq_idx, NULL, DOMID_SELF);
267 if (__predict_true(gdtset))
268 for (i = _xpq_idx; i > 0;) {
270 CTR6(KTR_PMAP, "mmu:val: %lx ptr: %lx val: %lx "
271 "ptr: %lx val: %lx ptr: %lx",
272 (XPQ_QUEUE[i-1].val & 0xffffffff),
273 (XPQ_QUEUE[i-1].ptr & 0xffffffff),
274 (XPQ_QUEUE[i-2].val & 0xffffffff),
275 (XPQ_QUEUE[i-2].ptr & 0xffffffff),
276 (XPQ_QUEUE[i-3].val & 0xffffffff),
277 (XPQ_QUEUE[i-3].ptr & 0xffffffff));
280 CTR4(KTR_PMAP, "mmu: val: %lx ptr: %lx val: %lx ptr: %lx",
281 (XPQ_QUEUE[i-1].val & 0xffffffff),
282 (XPQ_QUEUE[i-1].ptr & 0xffffffff),
283 (XPQ_QUEUE[i-2].val & 0xffffffff),
284 (XPQ_QUEUE[i-2].ptr & 0xffffffff));
287 CTR2(KTR_PMAP, "mmu: val: %lx ptr: %lx",
288 (XPQ_QUEUE[i-1].val & 0xffffffff),
289 (XPQ_QUEUE[i-1].ptr & 0xffffffff));
294 if (__predict_false(error < 0)) {
295 for (i = 0; i < _xpq_idx; i++)
296 printf("val: %llx ptr: %llx\n",
297 XPQ_QUEUE[i].val, XPQ_QUEUE[i].ptr);
298 panic("Failed to execute MMU updates: %d", error);
304 xen_flush_queue(void)
308 if (__predict_true(gdtset))
310 if (XPQ_IDX != 0) _xen_flush_queue();
311 if (__predict_true(gdtset))
316 xen_increment_idx(void)
321 if (__predict_false(XPQ_IDX == XPQUEUE_SIZE))
326 xen_check_queue(void)
331 KASSERT(XPQ_IDX == 0, ("pending operations XPQ_IDX=%d", XPQ_IDX));
336 xen_invlpg(vm_offset_t va)
339 op.cmd = MMUEXT_INVLPG_ALL;
340 op.arg1.linear_addr = va & ~PAGE_MASK;
341 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
345 xen_load_cr3(u_int val)
351 KASSERT(XPQ_IDX == 0, ("pending operations XPQ_IDX=%d", XPQ_IDX));
353 op.cmd = MMUEXT_NEW_BASEPTR;
354 op.arg1.mfn = xpmap_ptom(val) >> PAGE_SHIFT;
355 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
359 static __inline u_int
364 __asm __volatile("movl 4(%%ebp),%0" : "=r" (data));
375 eflags = _read_eflags();
376 _vcpu = &HYPERVISOR_shared_info->vcpu_info[smp_processor_id()];
377 if (_vcpu->evtchn_upcall_mask)
384 write_eflags(u_int eflags)
388 CTR2(KTR_SPARE2, "%x xen_restore_flags eflags %x", rebp(), eflags);
389 intr = ((eflags & PSL_I) == 0);
390 __restore_flags(intr);
391 _write_eflags(eflags);
397 CTR1(KTR_SPARE2, "%x xen_cli disabling interrupts", rebp());
404 CTR1(KTR_SPARE2, "%x xen_sti enabling interrupts", rebp());
412 return (HYPERVISOR_shared_info->vcpu_info[curcpu].arch.cr2);
416 _xen_machphys_update(vm_paddr_t mfn, vm_paddr_t pfn, char *file, int line)
420 if (__predict_true(gdtset))
422 XPQ_QUEUE[XPQ_IDX].ptr = (mfn << PAGE_SHIFT) | MMU_MACHPHYS_UPDATE;
423 XPQ_QUEUE[XPQ_IDX].val = pfn;
425 XPQ_QUEUE_LOG[XPQ_IDX].file = file;
426 XPQ_QUEUE_LOG[XPQ_IDX].line = line;
429 if (__predict_true(gdtset))
434 _xen_queue_pt_update(vm_paddr_t ptr, vm_paddr_t val, char *file, int line)
438 if (__predict_true(gdtset))
439 mtx_assert(&vm_page_queue_mtx, MA_OWNED);
441 KASSERT((ptr & 7) == 0, ("misaligned update"));
443 if (__predict_true(gdtset))
446 XPQ_QUEUE[XPQ_IDX].ptr = ((uint64_t)ptr) | MMU_NORMAL_PT_UPDATE;
447 XPQ_QUEUE[XPQ_IDX].val = (uint64_t)val;
449 XPQ_QUEUE_LOG[XPQ_IDX].file = file;
450 XPQ_QUEUE_LOG[XPQ_IDX].line = line;
453 if (__predict_true(gdtset))
458 xen_pgdpt_pin(vm_paddr_t ma)
461 op.cmd = MMUEXT_PIN_L3_TABLE;
462 op.arg1.mfn = ma >> PAGE_SHIFT;
464 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
468 xen_pgd_pin(vm_paddr_t ma)
471 op.cmd = MMUEXT_PIN_L2_TABLE;
472 op.arg1.mfn = ma >> PAGE_SHIFT;
474 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
478 xen_pgd_unpin(vm_paddr_t ma)
481 op.cmd = MMUEXT_UNPIN_TABLE;
482 op.arg1.mfn = ma >> PAGE_SHIFT;
484 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
488 xen_pt_pin(vm_paddr_t ma)
491 op.cmd = MMUEXT_PIN_L1_TABLE;
492 op.arg1.mfn = ma >> PAGE_SHIFT;
494 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
498 xen_pt_unpin(vm_paddr_t ma)
501 op.cmd = MMUEXT_UNPIN_TABLE;
502 op.arg1.mfn = ma >> PAGE_SHIFT;
504 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
508 xen_set_ldt(vm_paddr_t ptr, unsigned long len)
511 op.cmd = MMUEXT_SET_LDT;
512 op.arg1.linear_addr = ptr;
513 op.arg2.nr_ents = len;
515 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
518 void xen_tlb_flush(void)
521 op.cmd = MMUEXT_TLB_FLUSH_LOCAL;
523 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
527 xen_update_descriptor(union descriptor *table, union descriptor *entry)
532 ptp = vtopte((vm_offset_t)table);
533 pa = (*ptp & PG_FRAME) | ((vm_offset_t)table & PAGE_MASK);
534 if (HYPERVISOR_update_descriptor(pa, *(uint64_t *)entry))
535 panic("HYPERVISOR_update_descriptor failed\n");
541 * Bitmap is indexed by page number. If bit is set, the page is part of a
542 * xen_create_contiguous_region() area of memory.
544 unsigned long *contiguous_bitmap;
547 contiguous_bitmap_set(unsigned long first_page, unsigned long nr_pages)
549 unsigned long start_off, end_off, curr_idx, end_idx;
551 curr_idx = first_page / BITS_PER_LONG;
552 start_off = first_page & (BITS_PER_LONG-1);
553 end_idx = (first_page + nr_pages) / BITS_PER_LONG;
554 end_off = (first_page + nr_pages) & (BITS_PER_LONG-1);
556 if (curr_idx == end_idx) {
557 contiguous_bitmap[curr_idx] |=
558 ((1UL<<end_off)-1) & -(1UL<<start_off);
560 contiguous_bitmap[curr_idx] |= -(1UL<<start_off);
561 while ( ++curr_idx < end_idx )
562 contiguous_bitmap[curr_idx] = ~0UL;
563 contiguous_bitmap[curr_idx] |= (1UL<<end_off)-1;
568 contiguous_bitmap_clear(unsigned long first_page, unsigned long nr_pages)
570 unsigned long start_off, end_off, curr_idx, end_idx;
572 curr_idx = first_page / BITS_PER_LONG;
573 start_off = first_page & (BITS_PER_LONG-1);
574 end_idx = (first_page + nr_pages) / BITS_PER_LONG;
575 end_off = (first_page + nr_pages) & (BITS_PER_LONG-1);
577 if (curr_idx == end_idx) {
578 contiguous_bitmap[curr_idx] &=
579 -(1UL<<end_off) | ((1UL<<start_off)-1);
581 contiguous_bitmap[curr_idx] &= (1UL<<start_off)-1;
582 while ( ++curr_idx != end_idx )
583 contiguous_bitmap[curr_idx] = 0;
584 contiguous_bitmap[curr_idx] &= -(1UL<<end_off);
589 /* Ensure multi-page extents are contiguous in machine memory. */
591 xen_create_contiguous_region(vm_page_t pages, int npages)
593 unsigned long mfn, i, flags;
595 struct xen_memory_reservation reservation = {
600 set_xen_guest_handle(reservation.extent_start, &mfn);
604 /* can currently only handle power of two allocation */
605 PANIC_IF(ffs(npages) != fls(npages));
607 /* 0. determine order */
608 order = (ffs(npages) == fls(npages)) ? fls(npages) - 1 : fls(npages);
610 /* 1. give away machine pages. */
611 for (i = 0; i < (1 << order); i++) {
613 pfn = VM_PAGE_TO_PHYS(&pages[i]) >> PAGE_SHIFT;
615 PFNTOMFN(pfn) = INVALID_P2M_ENTRY;
616 PANIC_IF(HYPERVISOR_memory_op(XENMEM_decrease_reservation, &reservation) != 1);
620 /* 2. Get a new contiguous memory extent. */
621 reservation.extent_order = order;
622 /* xenlinux hardcodes this because of aacraid - maybe set to 0 if we're not
623 * running with a broxen driver XXXEN
625 reservation.address_bits = 31;
626 if (HYPERVISOR_memory_op(XENMEM_increase_reservation, &reservation) != 1)
629 /* 3. Map the new extent in place of old pages. */
630 for (i = 0; i < (1 << order); i++) {
632 pfn = VM_PAGE_TO_PHYS(&pages[i]) >> PAGE_SHIFT;
633 xen_machphys_update(mfn+i, pfn);
634 PFNTOMFN(pfn) = mfn+i;
640 contiguous_bitmap_set(VM_PAGE_TO_PHYS(&pages[0]) >> PAGE_SHIFT, 1UL << order);
643 balloon_unlock(flags);
648 reservation.extent_order = 0;
649 reservation.address_bits = 0;
651 for (i = 0; i < (1 << order); i++) {
653 pfn = VM_PAGE_TO_PHYS(&pages[i]) >> PAGE_SHIFT;
654 PANIC_IF(HYPERVISOR_memory_op(
655 XENMEM_increase_reservation, &reservation) != 1);
656 xen_machphys_update(mfn, pfn);
662 balloon_unlock(flags);
668 xen_destroy_contiguous_region(void *addr, int npages)
670 unsigned long mfn, i, flags, order, pfn0;
671 struct xen_memory_reservation reservation = {
676 set_xen_guest_handle(reservation.extent_start, &mfn);
678 pfn0 = vtophys(addr) >> PAGE_SHIFT;
680 scrub_pages(vstart, 1 << order);
682 /* can currently only handle power of two allocation */
683 PANIC_IF(ffs(npages) != fls(npages));
685 /* 0. determine order */
686 order = (ffs(npages) == fls(npages)) ? fls(npages) - 1 : fls(npages);
691 contiguous_bitmap_clear(vtophys(addr) >> PAGE_SHIFT, 1UL << order);
694 /* 1. Zap current PTEs, giving away the underlying pages. */
695 for (i = 0; i < (1 << order); i++) {
697 uint64_t new_val = 0;
698 pfn = vtomach((char *)addr + i*PAGE_SIZE) >> PAGE_SHIFT;
700 PANIC_IF(HYPERVISOR_update_va_mapping((vm_offset_t)((char *)addr + (i * PAGE_SIZE)), new_val, 0));
701 PFNTOMFN(pfn) = INVALID_P2M_ENTRY;
702 PANIC_IF(HYPERVISOR_memory_op(
703 XENMEM_decrease_reservation, &reservation) != 1);
706 /* 2. Map new pages in place of old pages. */
707 for (i = 0; i < (1 << order); i++) {
711 PANIC_IF(HYPERVISOR_memory_op(XENMEM_increase_reservation, &reservation) != 1);
713 new_val = mfn << PAGE_SHIFT;
714 PANIC_IF(HYPERVISOR_update_va_mapping((vm_offset_t)addr + (i * PAGE_SIZE),
715 new_val, PG_KERNEL));
716 xen_machphys_update(mfn, pfn);
722 balloon_unlock(flags);
725 extern vm_offset_t proc0kstack;
726 extern int vm86paddr, vm86phystk;
727 char *bootmem_start, *bootmem_current, *bootmem_end;
729 pteinfo_t *pteinfo_list;
730 void initvalues(start_info_t *startinfo);
732 struct xenstore_domain_interface;
733 extern struct xenstore_domain_interface *xen_store;
738 bootmem_alloc(unsigned int size)
742 retptr = bootmem_current;
743 PANIC_IF(retptr + size > bootmem_end);
744 bootmem_current += size;
750 bootmem_free(void *ptr, unsigned int size)
755 PANIC_IF(tptr != bootmem_current - size ||
756 bootmem_current - size < bootmem_start);
758 bootmem_current -= size;
763 xpmap_mtop2(vm_paddr_t mpa)
765 return ((machine_to_phys_mapping[mpa >> PAGE_SHIFT] << PAGE_SHIFT)
766 ) | (mpa & ~PG_FRAME);
770 xpmap_get_bootpde(vm_paddr_t va)
773 return ((pd_entry_t *)xen_start_info->pt_base)[va >> 22];
777 xpmap_get_vbootpde(vm_paddr_t va)
781 pde = xpmap_get_bootpde(va);
782 if ((pde & PG_V) == 0)
783 return (pde & ~PG_FRAME);
784 return (pde & ~PG_FRAME) |
785 (xpmap_mtop2(pde & PG_FRAME) + KERNBASE);
789 xpmap_get_bootptep(vm_paddr_t va)
793 pde = xpmap_get_vbootpde(va);
794 if ((pde & PG_V) == 0)
796 #define PT_MASK 0x003ff000 /* page table address bits */
797 return &(((pt_entry_t *)(pde & PG_FRAME))[(va & PT_MASK) >> PAGE_SHIFT]);
801 xpmap_get_bootpte(vm_paddr_t va)
804 return xpmap_get_bootptep(va)[0];
811 shift_phys_machine(unsigned long *phys_machine, int nr_pages)
814 unsigned long *tmp_page, *current_page, *next_page;
817 tmp_page = bootmem_alloc(PAGE_SIZE);
818 current_page = phys_machine + nr_pages - (PAGE_SIZE/sizeof(unsigned long));
819 next_page = current_page - (PAGE_SIZE/sizeof(unsigned long));
820 bcopy(phys_machine, tmp_page, PAGE_SIZE);
822 while (current_page > phys_machine) {
824 bcopy(next_page, tmp_page, PAGE_SIZE);
825 /* shift down page */
826 bcopy(current_page, next_page, PAGE_SIZE);
828 bcopy(tmp_page, current_page, PAGE_SIZE);
830 current_page -= (PAGE_SIZE/sizeof(unsigned long));
831 next_page -= (PAGE_SIZE/sizeof(unsigned long));
833 bootmem_free(tmp_page, PAGE_SIZE);
835 for (i = 0; i < nr_pages; i++) {
836 xen_machphys_update(phys_machine[i], i);
838 memset(phys_machine, INVALID_P2M_ENTRY, PAGE_SIZE);
841 #endif /* ADD_ISA_HOLE */
844 * Build a directory of the pages that make up our Physical to Machine
845 * mapping table. The Xen suspend/restore code uses this to find our
849 init_frame_list_list(void *arg)
851 unsigned long nr_pages = xen_start_info->nr_pages;
852 #define FPP (PAGE_SIZE/sizeof(xen_pfn_t))
855 xen_pfn_to_mfn_frame_list_list = malloc(PAGE_SIZE, M_DEVBUF, M_WAITOK);
856 for (i = 0, j = 0, k = -1; i < nr_pages;
858 if ((j & (FPP - 1)) == 0) {
860 xen_pfn_to_mfn_frame_list[k] =
861 malloc(PAGE_SIZE, M_DEVBUF, M_WAITOK);
862 xen_pfn_to_mfn_frame_list_list[k] =
863 VTOMFN(xen_pfn_to_mfn_frame_list[k]);
866 xen_pfn_to_mfn_frame_list[k][j] =
867 VTOMFN(&xen_phys_machine[i]);
870 HYPERVISOR_shared_info->arch.max_pfn = nr_pages;
871 HYPERVISOR_shared_info->arch.pfn_to_mfn_frame_list_list
872 = VTOMFN(xen_pfn_to_mfn_frame_list_list);
874 SYSINIT(init_fll, SI_SUB_DEVFS, SI_ORDER_ANY, init_frame_list_list, NULL);
876 extern unsigned long physfree;
881 extern uint32_t kernbase;
884 initvalues(start_info_t *startinfo)
886 vm_offset_t cur_space, cur_space_pt;
887 struct physdev_set_iopl set_iopl;
889 int l3_pages, l2_pages, l1_pages, offset;
890 vm_paddr_t console_page_ma, xen_store_ma;
894 vm_paddr_t IdlePDPTma, IdlePDPTnewma;
895 vm_paddr_t IdlePTDnewma[4];
896 pd_entry_t *IdlePDPTnew, *IdlePTDnew;
897 vm_paddr_t IdlePTDma[4];
899 vm_paddr_t IdlePTDma[1];
906 max((startinfo->nr_pages >> NPGPTD_SHIFT), nkpt),
907 NPGPTD*NPDEPG - KPTDI),
908 (HYPERVISOR_VIRT_START - KERNBASE) >> PDRSHIFT);
910 HYPERVISOR_vm_assist(VMASST_CMD_enable, VMASST_TYPE_4gb_segments);
913 * need to install handler
915 HYPERVISOR_vm_assist(VMASST_CMD_enable, VMASST_TYPE_4gb_segments_notify);
917 xen_start_info = startinfo;
918 xen_phys_machine = (xen_pfn_t *)startinfo->mfn_list;
920 IdlePTD = (pd_entry_t *)((uint8_t *)startinfo->pt_base + PAGE_SIZE);
926 IdlePDPT = (pd_entry_t *)startinfo->pt_base;
927 IdlePDPTma = VTOM(startinfo->pt_base);
928 for (i = (KERNBASE >> 30);
929 (i < 4) && (IdlePDPT[i] != 0); i++)
932 * Note that only one page directory has been allocated at this point.
935 for (i = 0; i < l2_pages; i++)
936 IdlePTDma[i] = VTOM(IdlePTD + i*PAGE_SIZE);
938 l2_pages = (l2_pages == 0) ? 1 : l2_pages;
943 for (i = (((KERNBASE>>18) & PAGE_MASK)>>PAGE_SHIFT);
944 (i<l2_pages*NPDEPG) && (i<(VM_MAX_KERNEL_ADDRESS>>PDRSHIFT)); i++) {
951 /* number of pages allocated after the pts + 1*/;
952 cur_space = xen_start_info->pt_base +
953 (l3_pages + l2_pages + l1_pages + 1)*PAGE_SIZE;
955 printk("initvalues(): wooh - availmem=%x,%x\n", avail_space, cur_space);
957 printk("KERNBASE=%x,pt_base=%x, VTOPFN(base)=%x, nr_pt_frames=%x\n",
958 KERNBASE,xen_start_info->pt_base, VTOPFN(xen_start_info->pt_base),
959 xen_start_info->nr_pt_frames);
960 xendebug_flags = 0; /* 0xffffffff; */
963 shift_phys_machine(xen_phys_machine, xen_start_info->nr_pages);
965 XENPRINTF("IdlePTD %p\n", IdlePTD);
966 XENPRINTF("nr_pages: %ld shared_info: 0x%lx flags: 0x%lx pt_base: 0x%lx "
967 "mod_start: 0x%lx mod_len: 0x%lx\n",
968 xen_start_info->nr_pages, xen_start_info->shared_info,
969 xen_start_info->flags, xen_start_info->pt_base,
970 xen_start_info->mod_start, xen_start_info->mod_len);
973 IdlePDPTnew = (pd_entry_t *)cur_space; cur_space += PAGE_SIZE;
974 bzero(IdlePDPTnew, PAGE_SIZE);
976 IdlePDPTnewma = VTOM(IdlePDPTnew);
977 IdlePTDnew = (pd_entry_t *)cur_space; cur_space += 4*PAGE_SIZE;
978 bzero(IdlePTDnew, 4*PAGE_SIZE);
980 for (i = 0; i < 4; i++)
981 IdlePTDnewma[i] = VTOM((uint8_t *)IdlePTDnew + i*PAGE_SIZE);
985 * Copy the 4 machine addresses of the new PTDs in to the PDPT
988 for (i = 0; i < 4; i++)
989 IdlePDPTnew[i] = IdlePTDnewma[i] | PG_V;
994 * re-map the new PDPT read-only
996 PT_SET_MA(IdlePDPTnew, IdlePDPTnewma | PG_V);
999 * Unpin the current PDPT
1001 xen_pt_unpin(IdlePDPTma);
1005 /* Map proc0's KSTACK */
1006 proc0kstack = cur_space; cur_space += (KSTACK_PAGES * PAGE_SIZE);
1007 printk("proc0kstack=%u\n", proc0kstack);
1009 /* vm86/bios stack */
1010 cur_space += PAGE_SIZE;
1012 /* Map space for the vm86 region */
1013 vm86paddr = (vm_offset_t)cur_space;
1014 cur_space += (PAGE_SIZE * 3);
1016 /* allocate 4 pages for bootmem allocator */
1017 bootmem_start = bootmem_current = (char *)cur_space;
1018 cur_space += (4 * PAGE_SIZE);
1019 bootmem_end = (char *)cur_space;
1021 /* allocate pages for gdt */
1022 gdt = (union descriptor *)cur_space;
1023 cur_space += PAGE_SIZE*ncpus;
1025 /* allocate page for ldt */
1026 ldt = (union descriptor *)cur_space; cur_space += PAGE_SIZE;
1027 cur_space += PAGE_SIZE;
1029 /* unmap remaining pages from initial chunk
1032 for (tmpva = cur_space; tmpva < (((uint32_t)&kernbase) + (l1_pages<<PDRSHIFT));
1033 tmpva += PAGE_SIZE) {
1034 bzero((char *)tmpva, PAGE_SIZE);
1035 PT_SET_MA(tmpva, (vm_paddr_t)0);
1040 memcpy(((uint8_t *)IdlePTDnew) + ((unsigned int)(KERNBASE >> 18)),
1041 ((uint8_t *)IdlePTD) + ((KERNBASE >> 18) & PAGE_MASK),
1042 l1_pages*sizeof(pt_entry_t));
1044 for (i = 0; i < 4; i++) {
1045 PT_SET_MA((uint8_t *)IdlePTDnew + i*PAGE_SIZE,
1046 IdlePTDnewma[i] | PG_V);
1048 xen_load_cr3(VTOP(IdlePDPTnew));
1049 xen_pgdpt_pin(VTOM(IdlePDPTnew));
1051 /* allocate remainder of nkpt pages */
1052 cur_space_pt = cur_space;
1053 for (offset = (KERNBASE >> PDRSHIFT), i = l1_pages; i < nkpt;
1054 i++, cur_space += PAGE_SIZE) {
1055 pdir = (offset + i) / NPDEPG;
1056 curoffset = ((offset + i) % NPDEPG);
1057 if (((offset + i) << PDRSHIFT) == VM_MAX_KERNEL_ADDRESS)
1061 * make sure that all the initial page table pages
1064 PT_SET_MA(cur_space, VTOM(cur_space) | PG_V | PG_RW);
1065 bzero((char *)cur_space, PAGE_SIZE);
1066 PT_SET_MA(cur_space, (vm_paddr_t)0);
1067 xen_pt_pin(VTOM(cur_space));
1068 xen_queue_pt_update((vm_paddr_t)(IdlePTDnewma[pdir] +
1069 curoffset*sizeof(vm_paddr_t)),
1070 VTOM(cur_space) | PG_KERNEL);
1074 for (i = 0; i < 4; i++) {
1075 pdir = (PTDPTDI + i) / NPDEPG;
1076 curoffset = (PTDPTDI + i) % NPDEPG;
1078 xen_queue_pt_update((vm_paddr_t)(IdlePTDnewma[pdir] +
1079 curoffset*sizeof(vm_paddr_t)),
1080 IdlePTDnewma[i] | PG_V);
1085 IdlePTD = IdlePTDnew;
1086 IdlePDPT = IdlePDPTnew;
1087 IdlePDPTma = IdlePDPTnewma;
1089 HYPERVISOR_shared_info = (shared_info_t *)cur_space;
1090 cur_space += PAGE_SIZE;
1092 xen_store = (struct xenstore_domain_interface *)cur_space;
1093 cur_space += PAGE_SIZE;
1095 console_page = (char *)cur_space;
1096 cur_space += PAGE_SIZE;
1099 * shared_info is an unsigned long so this will randomly break if
1100 * it is allocated above 4GB - I guess people are used to that
1101 * sort of thing with Xen ... sigh
1103 shinfo = xen_start_info->shared_info;
1104 PT_SET_MA(HYPERVISOR_shared_info, shinfo | PG_KERNEL);
1108 xen_store_ma = (((vm_paddr_t)xen_start_info->store_mfn) << PAGE_SHIFT);
1109 PT_SET_MA(xen_store, xen_store_ma | PG_KERNEL);
1110 console_page_ma = (((vm_paddr_t)xen_start_info->console.domU.mfn) << PAGE_SHIFT);
1111 PT_SET_MA(console_page, console_page_ma | PG_KERNEL);
1116 PANIC_IF(HYPERVISOR_physdev_op(PHYSDEVOP_SET_IOPL, &set_iopl));
1119 /* add page table for KERNBASE */
1120 xen_queue_pt_update(IdlePTDma + KPTDI*sizeof(vm_paddr_t),
1121 VTOM(cur_space) | PG_KERNEL);
1124 xen_queue_pt_update(pdir_shadow_ma[3] + KPTDI*sizeof(vm_paddr_t),
1125 VTOM(cur_space) | PG_V | PG_A);
1127 xen_queue_pt_update(pdir_shadow_ma + KPTDI*sizeof(vm_paddr_t),
1128 VTOM(cur_space) | PG_V | PG_A);
1131 cur_space += PAGE_SIZE;
1135 if (xen_start_info->flags & SIF_INITDOMAIN) {
1136 /* Map first megabyte */
1137 for (i = 0; i < (256 << PAGE_SHIFT); i += PAGE_SIZE)
1138 PT_SET_MA(KERNBASE + i, i | PG_KERNEL | PG_NC_PCD);
1143 * re-map kernel text read-only
1146 for (i = (((vm_offset_t)&btext) & ~PAGE_MASK);
1147 i < (((vm_offset_t)&etext) & ~PAGE_MASK); i += PAGE_SIZE)
1148 PT_SET_MA(i, VTOM(i) | PG_V | PG_A);
1151 physfree = VTOP(cur_space);
1152 init_first = physfree >> PAGE_SHIFT;
1153 IdlePTD = (pd_entry_t *)VTOP(IdlePTD);
1154 IdlePDPT = (pd_entry_t *)VTOP(IdlePDPT);
1155 setup_xen_features();
1156 printk("#8, proc0kstack=%u\n", proc0kstack);
1160 trap_info_t trap_table[] = {
1161 { 0, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(div)},
1162 { 1, 0|4, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(dbg)},
1163 { 3, 3|4, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(bpt)},
1164 { 4, 3, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(ofl)},
1165 /* This is UPL on Linux and KPL on BSD */
1166 { 5, 3, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(bnd)},
1167 { 6, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(ill)},
1168 { 7, 0|4, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(dna)},
1170 * { 8, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(XXX)},
1171 * no handler for double fault
1173 { 9, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(fpusegm)},
1174 {10, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(tss)},
1175 {11, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(missing)},
1176 {12, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(stk)},
1177 {13, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(prot)},
1178 {14, 0|4, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(page)},
1179 {15, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(rsvd)},
1180 {16, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(fpu)},
1181 {17, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(align)},
1182 {18, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(mchk)},
1183 {19, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(xmm)},
1184 {0x80, 3, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(int0x80_syscall)},
1188 /* Perform a multicall and check that individual calls succeeded. */
1190 HYPERVISOR_multicall(struct multicall_entry * call_list, int nr_calls)
1195 /* Perform the multicall. */
1196 PANIC_IF(_HYPERVISOR_multicall(call_list, nr_calls));
1198 /* Check the results of individual hypercalls. */
1199 for (i = 0; i < nr_calls; i++)
1200 if (unlikely(call_list[i].result < 0))
1202 if (unlikely(ret > 0))
1203 panic("%d multicall(s) failed: cpu %d\n",
1204 ret, smp_processor_id());
1206 /* If we didn't panic already, everything succeeded. */
1210 /********** CODE WORTH KEEPING ABOVE HERE *****************/
1212 void xen_failsafe_handler(void);
1215 xen_failsafe_handler(void)
1218 panic("xen_failsafe_handler called!\n");
1221 void xen_handle_thread_switch(struct pcb *pcb);
1223 /* This is called by cpu_switch() when switching threads. */
1224 /* The pcb arg refers to the process control block of the */
1225 /* next thread which is to run */
1227 xen_handle_thread_switch(struct pcb *pcb)
1229 uint32_t *a = (uint32_t *)&PCPU_GET(fsgs_gdt)[0];
1230 uint32_t *b = (uint32_t *)&pcb->pcb_fsd;
1231 multicall_entry_t mcl[3];
1234 /* Notify Xen of task switch */
1235 mcl[i].op = __HYPERVISOR_stack_switch;
1236 mcl[i].args[0] = GSEL(GDATA_SEL, SEL_KPL);
1237 mcl[i++].args[1] = (unsigned long)pcb;
1239 /* Check for update of fsd */
1240 if (*a != *b || *(a+1) != *(b+1)) {
1241 mcl[i].op = __HYPERVISOR_update_descriptor;
1242 *(uint64_t *)&mcl[i].args[0] = vtomach((vm_offset_t)a);
1243 *(uint64_t *)&mcl[i++].args[2] = *(uint64_t *)b;
1249 /* Check for update of gsd */
1250 if (*a != *b || *(a+1) != *(b+1)) {
1251 mcl[i].op = __HYPERVISOR_update_descriptor;
1252 *(uint64_t *)&mcl[i].args[0] = vtomach((vm_offset_t)a);
1253 *(uint64_t *)&mcl[i++].args[2] = *(uint64_t *)b;
1256 (void)HYPERVISOR_multicall(mcl, i);