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/rwlock.h>
48 #include <sys/sysproto.h>
51 #include <xen/xen-os.h>
55 #include <machine/segments.h>
56 #include <machine/pcb.h>
57 #include <machine/stdarg.h>
58 #include <machine/vmparam.h>
59 #include <machine/cpu.h>
60 #include <machine/intr_machdep.h>
61 #include <machine/md_var.h>
62 #include <machine/asmacros.h>
66 #include <xen/hypervisor.h>
67 #include <xen/xenstore/xenstorevar.h>
68 #include <machine/xen/xenvar.h>
69 #include <machine/xen/xenfunc.h>
70 #include <machine/xen/xenpmap.h>
71 #include <machine/xen/xenfunc.h>
72 #include <xen/interface/memory.h>
73 #include <machine/xen/features.h>
75 #include <machine/privatespace.h>
79 #include <vm/vm_page.h>
82 #define IDTVEC(name) __CONCAT(X,name)
85 IDTVEC(div), IDTVEC(dbg), IDTVEC(nmi), IDTVEC(bpt), IDTVEC(ofl),
86 IDTVEC(bnd), IDTVEC(ill), IDTVEC(dna), IDTVEC(fpusegm),
87 IDTVEC(tss), IDTVEC(missing), IDTVEC(stk), IDTVEC(prot),
88 IDTVEC(page), IDTVEC(mchk), IDTVEC(rsvd), IDTVEC(fpu), IDTVEC(align),
89 IDTVEC(xmm), IDTVEC(lcall_syscall), IDTVEC(int0x80_syscall);
93 start_info_t *xen_start_info;
94 start_info_t *HYPERVISOR_start_info;
95 shared_info_t *HYPERVISOR_shared_info;
96 xen_pfn_t *xen_machine_phys = machine_to_phys_mapping;
97 xen_pfn_t *xen_phys_machine;
98 xen_pfn_t *xen_pfn_to_mfn_frame_list[16];
99 xen_pfn_t *xen_pfn_to_mfn_frame_list_list;
100 int preemptable, init_first;
101 extern unsigned int avail_space;
102 int xen_vector_callback_enabled = 0;
103 enum xen_domain_type xen_domain_type = XEN_PV_DOMAIN;
112 CTR0(KTR_SPARE2, "ni_cli disabling interrupts");
113 __asm__("pushl %edx;"
126 __asm__("pushl %edx;"
138 force_evtchn_callback(void)
140 (void)HYPERVISOR_xen_version(0, NULL);
144 * Modify the cmd_line by converting ',' to NULLs so that it is in a format
145 * suitable for the static env vars.
148 xen_setbootenv(char *cmd_line)
152 /* Skip leading spaces */
153 for (; *cmd_line == ' '; cmd_line++);
155 xc_printf("xen_setbootenv(): cmd_line='%s'\n", cmd_line);
157 for (cmd_line_next = cmd_line; strsep(&cmd_line_next, ",") != NULL;);
162 xen_boothowto(char *envp)
166 /* get equivalents from the environment */
167 for (i = 0; howto_names[i].ev != NULL; i++)
168 if (getenv(howto_names[i].ev) != NULL)
169 howto |= howto_names[i].mask;
174 #define XPQUEUE_SIZE 128
182 /* per-cpu queues and indices */
184 static struct mmu_log xpq_queue_log[XEN_LEGACY_MAX_VCPUS][XPQUEUE_SIZE];
187 static int xpq_idx[XEN_LEGACY_MAX_VCPUS];
188 static mmu_update_t xpq_queue[XEN_LEGACY_MAX_VCPUS][XPQUEUE_SIZE];
190 #define XPQ_QUEUE_LOG xpq_queue_log[vcpu]
191 #define XPQ_QUEUE xpq_queue[vcpu]
192 #define XPQ_IDX xpq_idx[vcpu]
193 #define SET_VCPU() int vcpu = smp_processor_id()
196 static mmu_update_t xpq_queue[XPQUEUE_SIZE];
198 static struct mmu_log xpq_queue_log[XPQUEUE_SIZE];
200 static int xpq_idx = 0;
202 #define XPQ_QUEUE_LOG xpq_queue_log
203 #define XPQ_QUEUE xpq_queue
204 #define XPQ_IDX xpq_idx
208 #define XPQ_IDX_INC atomic_add_int(&XPQ_IDX, 1);
214 int _xpq_idx = XPQ_IDX;
220 xc_printf("xen_dump_queue(): %u entries\n", _xpq_idx);
221 for (i = 0; i < _xpq_idx; i++) {
222 xc_printf(" val: %llx ptr: %llx\n", XPQ_QUEUE[i].val,
230 _xen_flush_queue(void)
233 int _xpq_idx = XPQ_IDX;
237 if (__predict_true(gdtset))
238 CRITICAL_ASSERT(curthread);
242 /* Make sure index is cleared first to avoid double updates. */
243 error = HYPERVISOR_mmu_update((mmu_update_t *)&XPQ_QUEUE,
244 _xpq_idx, NULL, DOMID_SELF);
247 if (__predict_true(gdtset))
248 for (i = _xpq_idx; i > 0;) {
250 CTR6(KTR_PMAP, "mmu:val: %lx ptr: %lx val: %lx "
251 "ptr: %lx val: %lx ptr: %lx",
252 (XPQ_QUEUE[i-1].val & 0xffffffff),
253 (XPQ_QUEUE[i-1].ptr & 0xffffffff),
254 (XPQ_QUEUE[i-2].val & 0xffffffff),
255 (XPQ_QUEUE[i-2].ptr & 0xffffffff),
256 (XPQ_QUEUE[i-3].val & 0xffffffff),
257 (XPQ_QUEUE[i-3].ptr & 0xffffffff));
260 CTR4(KTR_PMAP, "mmu: val: %lx ptr: %lx val: %lx ptr: %lx",
261 (XPQ_QUEUE[i-1].val & 0xffffffff),
262 (XPQ_QUEUE[i-1].ptr & 0xffffffff),
263 (XPQ_QUEUE[i-2].val & 0xffffffff),
264 (XPQ_QUEUE[i-2].ptr & 0xffffffff));
267 CTR2(KTR_PMAP, "mmu: val: %lx ptr: %lx",
268 (XPQ_QUEUE[i-1].val & 0xffffffff),
269 (XPQ_QUEUE[i-1].ptr & 0xffffffff));
274 if (__predict_false(error < 0)) {
275 for (i = 0; i < _xpq_idx; i++)
276 printf("val: %llx ptr: %llx\n",
277 XPQ_QUEUE[i].val, XPQ_QUEUE[i].ptr);
278 panic("Failed to execute MMU updates: %d", error);
284 xen_flush_queue(void)
288 if (__predict_true(gdtset))
290 if (XPQ_IDX != 0) _xen_flush_queue();
291 if (__predict_true(gdtset))
296 xen_increment_idx(void)
301 if (__predict_false(XPQ_IDX == XPQUEUE_SIZE))
306 xen_check_queue(void)
311 KASSERT(XPQ_IDX == 0, ("pending operations XPQ_IDX=%d", XPQ_IDX));
316 xen_invlpg(vm_offset_t va)
319 op.cmd = MMUEXT_INVLPG_ALL;
320 op.arg1.linear_addr = va & ~PAGE_MASK;
321 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
325 xen_load_cr3(u_int val)
331 KASSERT(XPQ_IDX == 0, ("pending operations XPQ_IDX=%d", XPQ_IDX));
333 op.cmd = MMUEXT_NEW_BASEPTR;
334 op.arg1.mfn = xpmap_ptom(val) >> PAGE_SHIFT;
335 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
339 static __inline u_int
344 __asm __volatile("movl 4(%%ebp),%0" : "=r" (data));
355 eflags = _read_eflags();
356 _vcpu = &HYPERVISOR_shared_info->vcpu_info[smp_processor_id()];
357 if (_vcpu->evtchn_upcall_mask)
364 write_eflags(u_int eflags)
368 CTR2(KTR_SPARE2, "%x xen_restore_flags eflags %x", rebp(), eflags);
369 intr = ((eflags & PSL_I) == 0);
370 __restore_flags(intr);
371 _write_eflags(eflags);
377 CTR1(KTR_SPARE2, "%x xen_cli disabling interrupts", rebp());
384 CTR1(KTR_SPARE2, "%x xen_sti enabling interrupts", rebp());
392 return (HYPERVISOR_shared_info->vcpu_info[curcpu].arch.cr2);
396 _xen_machphys_update(vm_paddr_t mfn, vm_paddr_t pfn, char *file, int line)
400 if (__predict_true(gdtset))
402 XPQ_QUEUE[XPQ_IDX].ptr = (mfn << PAGE_SHIFT) | MMU_MACHPHYS_UPDATE;
403 XPQ_QUEUE[XPQ_IDX].val = pfn;
405 XPQ_QUEUE_LOG[XPQ_IDX].file = file;
406 XPQ_QUEUE_LOG[XPQ_IDX].line = line;
409 if (__predict_true(gdtset))
413 extern struct rwlock pvh_global_lock;
416 _xen_queue_pt_update(vm_paddr_t ptr, vm_paddr_t val, char *file, int line)
420 if (__predict_true(gdtset))
421 rw_assert(&pvh_global_lock, RA_WLOCKED);
423 KASSERT((ptr & 7) == 0, ("misaligned update"));
425 if (__predict_true(gdtset))
428 XPQ_QUEUE[XPQ_IDX].ptr = ((uint64_t)ptr) | MMU_NORMAL_PT_UPDATE;
429 XPQ_QUEUE[XPQ_IDX].val = (uint64_t)val;
431 XPQ_QUEUE_LOG[XPQ_IDX].file = file;
432 XPQ_QUEUE_LOG[XPQ_IDX].line = line;
435 if (__predict_true(gdtset))
440 xen_pgdpt_pin(vm_paddr_t ma)
443 op.cmd = MMUEXT_PIN_L3_TABLE;
444 op.arg1.mfn = ma >> PAGE_SHIFT;
446 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
450 xen_pgd_pin(vm_paddr_t ma)
453 op.cmd = MMUEXT_PIN_L2_TABLE;
454 op.arg1.mfn = ma >> PAGE_SHIFT;
456 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
460 xen_pgd_unpin(vm_paddr_t ma)
463 op.cmd = MMUEXT_UNPIN_TABLE;
464 op.arg1.mfn = ma >> PAGE_SHIFT;
466 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
470 xen_pt_pin(vm_paddr_t ma)
473 op.cmd = MMUEXT_PIN_L1_TABLE;
474 op.arg1.mfn = ma >> PAGE_SHIFT;
476 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
480 xen_pt_unpin(vm_paddr_t ma)
483 op.cmd = MMUEXT_UNPIN_TABLE;
484 op.arg1.mfn = ma >> PAGE_SHIFT;
486 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
490 xen_set_ldt(vm_paddr_t ptr, unsigned long len)
493 op.cmd = MMUEXT_SET_LDT;
494 op.arg1.linear_addr = ptr;
495 op.arg2.nr_ents = len;
497 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
500 void xen_tlb_flush(void)
503 op.cmd = MMUEXT_TLB_FLUSH_LOCAL;
505 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
509 xen_update_descriptor(union descriptor *table, union descriptor *entry)
514 ptp = vtopte((vm_offset_t)table);
515 pa = (*ptp & PG_FRAME) | ((vm_offset_t)table & PAGE_MASK);
516 if (HYPERVISOR_update_descriptor(pa, *(uint64_t *)entry))
517 panic("HYPERVISOR_update_descriptor failed\n");
523 * Bitmap is indexed by page number. If bit is set, the page is part of a
524 * xen_create_contiguous_region() area of memory.
526 unsigned long *contiguous_bitmap;
529 contiguous_bitmap_set(unsigned long first_page, unsigned long nr_pages)
531 unsigned long start_off, end_off, curr_idx, end_idx;
533 curr_idx = first_page / BITS_PER_LONG;
534 start_off = first_page & (BITS_PER_LONG-1);
535 end_idx = (first_page + nr_pages) / BITS_PER_LONG;
536 end_off = (first_page + nr_pages) & (BITS_PER_LONG-1);
538 if (curr_idx == end_idx) {
539 contiguous_bitmap[curr_idx] |=
540 ((1UL<<end_off)-1) & -(1UL<<start_off);
542 contiguous_bitmap[curr_idx] |= -(1UL<<start_off);
543 while ( ++curr_idx < end_idx )
544 contiguous_bitmap[curr_idx] = ~0UL;
545 contiguous_bitmap[curr_idx] |= (1UL<<end_off)-1;
550 contiguous_bitmap_clear(unsigned long first_page, unsigned long nr_pages)
552 unsigned long start_off, end_off, curr_idx, end_idx;
554 curr_idx = first_page / BITS_PER_LONG;
555 start_off = first_page & (BITS_PER_LONG-1);
556 end_idx = (first_page + nr_pages) / BITS_PER_LONG;
557 end_off = (first_page + nr_pages) & (BITS_PER_LONG-1);
559 if (curr_idx == end_idx) {
560 contiguous_bitmap[curr_idx] &=
561 -(1UL<<end_off) | ((1UL<<start_off)-1);
563 contiguous_bitmap[curr_idx] &= (1UL<<start_off)-1;
564 while ( ++curr_idx != end_idx )
565 contiguous_bitmap[curr_idx] = 0;
566 contiguous_bitmap[curr_idx] &= -(1UL<<end_off);
571 /* Ensure multi-page extents are contiguous in machine memory. */
573 xen_create_contiguous_region(vm_page_t pages, int npages)
575 unsigned long mfn, i, flags;
577 struct xen_memory_reservation reservation = {
582 set_xen_guest_handle(reservation.extent_start, &mfn);
586 /* can currently only handle power of two allocation */
587 PANIC_IF(ffs(npages) != fls(npages));
589 /* 0. determine order */
590 order = (ffs(npages) == fls(npages)) ? fls(npages) - 1 : fls(npages);
592 /* 1. give away machine pages. */
593 for (i = 0; i < (1 << order); i++) {
595 pfn = VM_PAGE_TO_PHYS(&pages[i]) >> PAGE_SHIFT;
597 PFNTOMFN(pfn) = INVALID_P2M_ENTRY;
598 PANIC_IF(HYPERVISOR_memory_op(XENMEM_decrease_reservation, &reservation) != 1);
602 /* 2. Get a new contiguous memory extent. */
603 reservation.extent_order = order;
604 /* xenlinux hardcodes this because of aacraid - maybe set to 0 if we're not
605 * running with a broxen driver XXXEN
607 reservation.address_bits = 31;
608 if (HYPERVISOR_memory_op(XENMEM_increase_reservation, &reservation) != 1)
611 /* 3. Map the new extent in place of old pages. */
612 for (i = 0; i < (1 << order); i++) {
614 pfn = VM_PAGE_TO_PHYS(&pages[i]) >> PAGE_SHIFT;
615 xen_machphys_update(mfn+i, pfn);
616 PFNTOMFN(pfn) = mfn+i;
622 contiguous_bitmap_set(VM_PAGE_TO_PHYS(&pages[0]) >> PAGE_SHIFT, 1UL << order);
625 balloon_unlock(flags);
630 reservation.extent_order = 0;
631 reservation.address_bits = 0;
633 for (i = 0; i < (1 << order); i++) {
635 pfn = VM_PAGE_TO_PHYS(&pages[i]) >> PAGE_SHIFT;
636 PANIC_IF(HYPERVISOR_memory_op(
637 XENMEM_increase_reservation, &reservation) != 1);
638 xen_machphys_update(mfn, pfn);
644 balloon_unlock(flags);
650 xen_destroy_contiguous_region(void *addr, int npages)
652 unsigned long mfn, i, flags, order, pfn0;
653 struct xen_memory_reservation reservation = {
658 set_xen_guest_handle(reservation.extent_start, &mfn);
660 pfn0 = vtophys(addr) >> PAGE_SHIFT;
662 scrub_pages(vstart, 1 << order);
664 /* can currently only handle power of two allocation */
665 PANIC_IF(ffs(npages) != fls(npages));
667 /* 0. determine order */
668 order = (ffs(npages) == fls(npages)) ? fls(npages) - 1 : fls(npages);
673 contiguous_bitmap_clear(vtophys(addr) >> PAGE_SHIFT, 1UL << order);
676 /* 1. Zap current PTEs, giving away the underlying pages. */
677 for (i = 0; i < (1 << order); i++) {
679 uint64_t new_val = 0;
680 pfn = vtomach((char *)addr + i*PAGE_SIZE) >> PAGE_SHIFT;
682 PANIC_IF(HYPERVISOR_update_va_mapping((vm_offset_t)((char *)addr + (i * PAGE_SIZE)), new_val, 0));
683 PFNTOMFN(pfn) = INVALID_P2M_ENTRY;
684 PANIC_IF(HYPERVISOR_memory_op(
685 XENMEM_decrease_reservation, &reservation) != 1);
688 /* 2. Map new pages in place of old pages. */
689 for (i = 0; i < (1 << order); i++) {
693 PANIC_IF(HYPERVISOR_memory_op(XENMEM_increase_reservation, &reservation) != 1);
695 new_val = mfn << PAGE_SHIFT;
696 PANIC_IF(HYPERVISOR_update_va_mapping((vm_offset_t)addr + (i * PAGE_SIZE),
697 new_val, PG_KERNEL));
698 xen_machphys_update(mfn, pfn);
704 balloon_unlock(flags);
707 extern vm_offset_t proc0kstack;
708 extern int vm86paddr, vm86phystk;
709 char *bootmem_start, *bootmem_current, *bootmem_end;
711 pteinfo_t *pteinfo_list;
712 void initvalues(start_info_t *startinfo);
715 bootmem_alloc(unsigned int size)
719 retptr = bootmem_current;
720 PANIC_IF(retptr + size > bootmem_end);
721 bootmem_current += size;
727 bootmem_free(void *ptr, unsigned int size)
732 PANIC_IF(tptr != bootmem_current - size ||
733 bootmem_current - size < bootmem_start);
735 bootmem_current -= size;
740 xpmap_mtop2(vm_paddr_t mpa)
742 return ((machine_to_phys_mapping[mpa >> PAGE_SHIFT] << PAGE_SHIFT)
743 ) | (mpa & ~PG_FRAME);
747 xpmap_get_bootpde(vm_paddr_t va)
750 return ((pd_entry_t *)xen_start_info->pt_base)[va >> 22];
754 xpmap_get_vbootpde(vm_paddr_t va)
758 pde = xpmap_get_bootpde(va);
759 if ((pde & PG_V) == 0)
760 return (pde & ~PG_FRAME);
761 return (pde & ~PG_FRAME) |
762 (xpmap_mtop2(pde & PG_FRAME) + KERNBASE);
766 xpmap_get_bootptep(vm_paddr_t va)
770 pde = xpmap_get_vbootpde(va);
771 if ((pde & PG_V) == 0)
773 #define PT_MASK 0x003ff000 /* page table address bits */
774 return &(((pt_entry_t *)(pde & PG_FRAME))[(va & PT_MASK) >> PAGE_SHIFT]);
778 xpmap_get_bootpte(vm_paddr_t va)
781 return xpmap_get_bootptep(va)[0];
788 shift_phys_machine(unsigned long *phys_machine, int nr_pages)
791 unsigned long *tmp_page, *current_page, *next_page;
794 tmp_page = bootmem_alloc(PAGE_SIZE);
795 current_page = phys_machine + nr_pages - (PAGE_SIZE/sizeof(unsigned long));
796 next_page = current_page - (PAGE_SIZE/sizeof(unsigned long));
797 bcopy(phys_machine, tmp_page, PAGE_SIZE);
799 while (current_page > phys_machine) {
801 bcopy(next_page, tmp_page, PAGE_SIZE);
802 /* shift down page */
803 bcopy(current_page, next_page, PAGE_SIZE);
805 bcopy(tmp_page, current_page, PAGE_SIZE);
807 current_page -= (PAGE_SIZE/sizeof(unsigned long));
808 next_page -= (PAGE_SIZE/sizeof(unsigned long));
810 bootmem_free(tmp_page, PAGE_SIZE);
812 for (i = 0; i < nr_pages; i++) {
813 xen_machphys_update(phys_machine[i], i);
815 memset(phys_machine, INVALID_P2M_ENTRY, PAGE_SIZE);
818 #endif /* ADD_ISA_HOLE */
821 * Build a directory of the pages that make up our Physical to Machine
822 * mapping table. The Xen suspend/restore code uses this to find our
826 init_frame_list_list(void *arg)
828 unsigned long nr_pages = xen_start_info->nr_pages;
829 #define FPP (PAGE_SIZE/sizeof(xen_pfn_t))
832 xen_pfn_to_mfn_frame_list_list = malloc(PAGE_SIZE, M_DEVBUF, M_WAITOK);
833 for (i = 0, j = 0, k = -1; i < nr_pages;
835 if ((j & (FPP - 1)) == 0) {
837 xen_pfn_to_mfn_frame_list[k] =
838 malloc(PAGE_SIZE, M_DEVBUF, M_WAITOK);
839 xen_pfn_to_mfn_frame_list_list[k] =
840 VTOMFN(xen_pfn_to_mfn_frame_list[k]);
843 xen_pfn_to_mfn_frame_list[k][j] =
844 VTOMFN(&xen_phys_machine[i]);
847 HYPERVISOR_shared_info->arch.max_pfn = nr_pages;
848 HYPERVISOR_shared_info->arch.pfn_to_mfn_frame_list_list
849 = VTOMFN(xen_pfn_to_mfn_frame_list_list);
851 SYSINIT(init_fll, SI_SUB_DEVFS, SI_ORDER_ANY, init_frame_list_list, NULL);
853 extern unsigned long physfree;
858 extern uint32_t kernbase;
861 initvalues(start_info_t *startinfo)
863 vm_offset_t cur_space, cur_space_pt;
864 struct physdev_set_iopl set_iopl;
866 int l3_pages, l2_pages, l1_pages, offset;
867 vm_paddr_t console_page_ma, xen_store_ma;
871 vm_paddr_t IdlePDPTma, IdlePDPTnewma;
872 vm_paddr_t IdlePTDnewma[4];
873 pd_entry_t *IdlePDPTnew, *IdlePTDnew;
874 vm_paddr_t IdlePTDma[4];
876 vm_paddr_t IdlePTDma[1];
883 max((startinfo->nr_pages >> NPGPTD_SHIFT), nkpt),
884 NPGPTD*NPDEPG - KPTDI),
885 (HYPERVISOR_VIRT_START - KERNBASE) >> PDRSHIFT);
887 HYPERVISOR_vm_assist(VMASST_CMD_enable, VMASST_TYPE_4gb_segments);
890 * need to install handler
892 HYPERVISOR_vm_assist(VMASST_CMD_enable, VMASST_TYPE_4gb_segments_notify);
894 xen_start_info = startinfo;
895 HYPERVISOR_start_info = startinfo;
896 xen_phys_machine = (xen_pfn_t *)startinfo->mfn_list;
898 IdlePTD = (pd_entry_t *)((uint8_t *)startinfo->pt_base + PAGE_SIZE);
904 IdlePDPT = (pd_entry_t *)startinfo->pt_base;
905 IdlePDPTma = VTOM(startinfo->pt_base);
906 for (i = (KERNBASE >> 30);
907 (i < 4) && (IdlePDPT[i] != 0); i++)
910 * Note that only one page directory has been allocated at this point.
913 for (i = 0; i < l2_pages; i++)
914 IdlePTDma[i] = VTOM(IdlePTD + i*PAGE_SIZE);
916 l2_pages = (l2_pages == 0) ? 1 : l2_pages;
921 for (i = (((KERNBASE>>18) & PAGE_MASK)>>PAGE_SHIFT);
922 (i<l2_pages*NPDEPG) && (i<(VM_MAX_KERNEL_ADDRESS>>PDRSHIFT)); i++) {
929 /* number of pages allocated after the pts + 1*/;
930 cur_space = xen_start_info->pt_base +
931 (l3_pages + l2_pages + l1_pages + 1)*PAGE_SIZE;
933 xc_printf("initvalues(): wooh - availmem=%x,%x\n", avail_space,
936 xc_printf("KERNBASE=%x,pt_base=%lx, VTOPFN(base)=%x, nr_pt_frames=%lx\n",
937 KERNBASE,xen_start_info->pt_base, VTOPFN(xen_start_info->pt_base),
938 xen_start_info->nr_pt_frames);
939 xendebug_flags = 0; /* 0xffffffff; */
942 shift_phys_machine(xen_phys_machine, xen_start_info->nr_pages);
944 XENPRINTF("IdlePTD %p\n", IdlePTD);
945 XENPRINTF("nr_pages: %ld shared_info: 0x%lx flags: 0x%x pt_base: 0x%lx "
946 "mod_start: 0x%lx mod_len: 0x%lx\n",
947 xen_start_info->nr_pages, xen_start_info->shared_info,
948 xen_start_info->flags, xen_start_info->pt_base,
949 xen_start_info->mod_start, xen_start_info->mod_len);
952 IdlePDPTnew = (pd_entry_t *)cur_space; cur_space += PAGE_SIZE;
953 bzero(IdlePDPTnew, PAGE_SIZE);
955 IdlePDPTnewma = VTOM(IdlePDPTnew);
956 IdlePTDnew = (pd_entry_t *)cur_space; cur_space += 4*PAGE_SIZE;
957 bzero(IdlePTDnew, 4*PAGE_SIZE);
959 for (i = 0; i < 4; i++)
960 IdlePTDnewma[i] = VTOM((uint8_t *)IdlePTDnew + i*PAGE_SIZE);
964 * Copy the 4 machine addresses of the new PTDs in to the PDPT
967 for (i = 0; i < 4; i++)
968 IdlePDPTnew[i] = IdlePTDnewma[i] | PG_V;
973 * re-map the new PDPT read-only
975 PT_SET_MA(IdlePDPTnew, IdlePDPTnewma | PG_V);
978 * Unpin the current PDPT
980 xen_pt_unpin(IdlePDPTma);
984 /* Map proc0's KSTACK */
985 proc0kstack = cur_space; cur_space += (KSTACK_PAGES * PAGE_SIZE);
986 xc_printf("proc0kstack=%u\n", proc0kstack);
988 /* vm86/bios stack */
989 cur_space += PAGE_SIZE;
991 /* Map space for the vm86 region */
992 vm86paddr = (vm_offset_t)cur_space;
993 cur_space += (PAGE_SIZE * 3);
995 /* allocate 4 pages for bootmem allocator */
996 bootmem_start = bootmem_current = (char *)cur_space;
997 cur_space += (4 * PAGE_SIZE);
998 bootmem_end = (char *)cur_space;
1000 /* allocate pages for gdt */
1001 gdt = (union descriptor *)cur_space;
1002 cur_space += PAGE_SIZE*ncpus;
1004 /* allocate page for ldt */
1005 ldt = (union descriptor *)cur_space; cur_space += PAGE_SIZE;
1006 cur_space += PAGE_SIZE;
1008 /* unmap remaining pages from initial chunk
1011 for (tmpva = cur_space; tmpva < (((uint32_t)&kernbase) + (l1_pages<<PDRSHIFT));
1012 tmpva += PAGE_SIZE) {
1013 bzero((char *)tmpva, PAGE_SIZE);
1014 PT_SET_MA(tmpva, (vm_paddr_t)0);
1019 memcpy(((uint8_t *)IdlePTDnew) + ((unsigned int)(KERNBASE >> 18)),
1020 ((uint8_t *)IdlePTD) + ((KERNBASE >> 18) & PAGE_MASK),
1021 l1_pages*sizeof(pt_entry_t));
1023 for (i = 0; i < 4; i++) {
1024 PT_SET_MA((uint8_t *)IdlePTDnew + i*PAGE_SIZE,
1025 IdlePTDnewma[i] | PG_V);
1027 xen_load_cr3(VTOP(IdlePDPTnew));
1028 xen_pgdpt_pin(VTOM(IdlePDPTnew));
1030 /* allocate remainder of nkpt pages */
1031 cur_space_pt = cur_space;
1032 for (offset = (KERNBASE >> PDRSHIFT), i = l1_pages; i < nkpt;
1033 i++, cur_space += PAGE_SIZE) {
1034 pdir = (offset + i) / NPDEPG;
1035 curoffset = ((offset + i) % NPDEPG);
1036 if (((offset + i) << PDRSHIFT) == VM_MAX_KERNEL_ADDRESS)
1040 * make sure that all the initial page table pages
1043 PT_SET_MA(cur_space, VTOM(cur_space) | PG_V | PG_RW);
1044 bzero((char *)cur_space, PAGE_SIZE);
1045 PT_SET_MA(cur_space, (vm_paddr_t)0);
1046 xen_pt_pin(VTOM(cur_space));
1047 xen_queue_pt_update((vm_paddr_t)(IdlePTDnewma[pdir] +
1048 curoffset*sizeof(vm_paddr_t)),
1049 VTOM(cur_space) | PG_KERNEL);
1053 for (i = 0; i < 4; i++) {
1054 pdir = (PTDPTDI + i) / NPDEPG;
1055 curoffset = (PTDPTDI + i) % NPDEPG;
1057 xen_queue_pt_update((vm_paddr_t)(IdlePTDnewma[pdir] +
1058 curoffset*sizeof(vm_paddr_t)),
1059 IdlePTDnewma[i] | PG_V);
1064 IdlePTD = IdlePTDnew;
1065 IdlePDPT = IdlePDPTnew;
1066 IdlePDPTma = IdlePDPTnewma;
1068 HYPERVISOR_shared_info = (shared_info_t *)cur_space;
1069 cur_space += PAGE_SIZE;
1071 xen_store = (struct xenstore_domain_interface *)cur_space;
1072 cur_space += PAGE_SIZE;
1074 console_page = (char *)cur_space;
1075 cur_space += PAGE_SIZE;
1078 * shared_info is an unsigned long so this will randomly break if
1079 * it is allocated above 4GB - I guess people are used to that
1080 * sort of thing with Xen ... sigh
1082 shinfo = xen_start_info->shared_info;
1083 PT_SET_MA(HYPERVISOR_shared_info, shinfo | PG_KERNEL);
1087 xen_store_ma = (((vm_paddr_t)xen_start_info->store_mfn) << PAGE_SHIFT);
1088 PT_SET_MA(xen_store, xen_store_ma | PG_KERNEL);
1089 console_page_ma = (((vm_paddr_t)xen_start_info->console.domU.mfn) << PAGE_SHIFT);
1090 PT_SET_MA(console_page, console_page_ma | PG_KERNEL);
1095 PANIC_IF(HYPERVISOR_physdev_op(PHYSDEVOP_SET_IOPL, &set_iopl));
1098 /* add page table for KERNBASE */
1099 xen_queue_pt_update(IdlePTDma + KPTDI*sizeof(vm_paddr_t),
1100 VTOM(cur_space) | PG_KERNEL);
1103 xen_queue_pt_update(pdir_shadow_ma[3] + KPTDI*sizeof(vm_paddr_t),
1104 VTOM(cur_space) | PG_V | PG_A);
1106 xen_queue_pt_update(pdir_shadow_ma + KPTDI*sizeof(vm_paddr_t),
1107 VTOM(cur_space) | PG_V | PG_A);
1110 cur_space += PAGE_SIZE;
1114 if (xen_start_info->flags & SIF_INITDOMAIN) {
1115 /* Map first megabyte */
1116 for (i = 0; i < (256 << PAGE_SHIFT); i += PAGE_SIZE)
1117 PT_SET_MA(KERNBASE + i, i | PG_KERNEL | PG_NC_PCD);
1122 * re-map kernel text read-only
1125 for (i = (((vm_offset_t)&btext) & ~PAGE_MASK);
1126 i < (((vm_offset_t)&etext) & ~PAGE_MASK); i += PAGE_SIZE)
1127 PT_SET_MA(i, VTOM(i) | PG_V | PG_A);
1130 physfree = VTOP(cur_space);
1131 init_first = physfree >> PAGE_SHIFT;
1132 IdlePTD = (pd_entry_t *)VTOP(IdlePTD);
1133 IdlePDPT = (pd_entry_t *)VTOP(IdlePDPT);
1134 setup_xen_features();
1135 xc_printf("#8, proc0kstack=%u\n", proc0kstack);
1139 trap_info_t trap_table[] = {
1140 { 0, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(div)},
1141 { 1, 0|4, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(dbg)},
1142 { 3, 3|4, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(bpt)},
1143 { 4, 3, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(ofl)},
1144 /* This is UPL on Linux and KPL on BSD */
1145 { 5, 3, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(bnd)},
1146 { 6, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(ill)},
1147 { 7, 0|4, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(dna)},
1149 * { 8, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(XXX)},
1150 * no handler for double fault
1152 { 9, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(fpusegm)},
1153 {10, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(tss)},
1154 {11, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(missing)},
1155 {12, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(stk)},
1156 {13, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(prot)},
1157 {14, 0|4, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(page)},
1158 {15, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(rsvd)},
1159 {16, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(fpu)},
1160 {17, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(align)},
1161 {18, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(mchk)},
1162 {19, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(xmm)},
1163 {0x80, 3, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(int0x80_syscall)},
1167 /* Perform a multicall and check that individual calls succeeded. */
1169 HYPERVISOR_multicall(struct multicall_entry * call_list, int nr_calls)
1174 /* Perform the multicall. */
1175 PANIC_IF(_HYPERVISOR_multicall(call_list, nr_calls));
1177 /* Check the results of individual hypercalls. */
1178 for (i = 0; i < nr_calls; i++)
1179 if (__predict_false(call_list[i].result < 0))
1181 if (__predict_false(ret > 0))
1182 panic("%d multicall(s) failed: cpu %d\n",
1183 ret, smp_processor_id());
1185 /* If we didn't panic already, everything succeeded. */
1189 /********** CODE WORTH KEEPING ABOVE HERE *****************/
1191 void xen_failsafe_handler(void);
1194 xen_failsafe_handler(void)
1197 panic("xen_failsafe_handler called!\n");
1200 void xen_handle_thread_switch(struct pcb *pcb);
1202 /* This is called by cpu_switch() when switching threads. */
1203 /* The pcb arg refers to the process control block of the */
1204 /* next thread which is to run */
1206 xen_handle_thread_switch(struct pcb *pcb)
1208 uint32_t *a = (uint32_t *)&PCPU_GET(fsgs_gdt)[0];
1209 uint32_t *b = (uint32_t *)&pcb->pcb_fsd;
1210 multicall_entry_t mcl[3];
1213 /* Notify Xen of task switch */
1214 mcl[i].op = __HYPERVISOR_stack_switch;
1215 mcl[i].args[0] = GSEL(GDATA_SEL, SEL_KPL);
1216 mcl[i++].args[1] = (unsigned long)pcb;
1218 /* Check for update of fsd */
1219 if (*a != *b || *(a+1) != *(b+1)) {
1220 mcl[i].op = __HYPERVISOR_update_descriptor;
1221 *(uint64_t *)&mcl[i].args[0] = vtomach((vm_offset_t)a);
1222 *(uint64_t *)&mcl[i++].args[2] = *(uint64_t *)b;
1228 /* Check for update of gsd */
1229 if (*a != *b || *(a+1) != *(b+1)) {
1230 mcl[i].op = __HYPERVISOR_update_descriptor;
1231 *(uint64_t *)&mcl[i].args[0] = vtomach((vm_offset_t)a);
1232 *(uint64_t *)&mcl[i++].args[2] = *(uint64_t *)b;
1235 (void)HYPERVISOR_multicall(mcl, i);