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>
50 #include <xen/xen-os.h>
54 #include <machine/segments.h>
55 #include <machine/pcb.h>
56 #include <machine/stdarg.h>
57 #include <machine/vmparam.h>
58 #include <machine/cpu.h>
59 #include <machine/intr_machdep.h>
60 #include <machine/md_var.h>
61 #include <machine/asmacros.h>
65 #include <xen/hypervisor.h>
66 #include <machine/xen/xenvar.h>
67 #include <machine/xen/xenfunc.h>
68 #include <machine/xen/xenpmap.h>
69 #include <machine/xen/xenfunc.h>
70 #include <xen/interface/memory.h>
71 #include <machine/xen/features.h>
73 #include <machine/privatespace.h>
77 #include <vm/vm_page.h>
80 #define IDTVEC(name) __CONCAT(X,name)
83 IDTVEC(div), IDTVEC(dbg), IDTVEC(nmi), IDTVEC(bpt), IDTVEC(ofl),
84 IDTVEC(bnd), IDTVEC(ill), IDTVEC(dna), IDTVEC(fpusegm),
85 IDTVEC(tss), IDTVEC(missing), IDTVEC(stk), IDTVEC(prot),
86 IDTVEC(page), IDTVEC(mchk), IDTVEC(rsvd), IDTVEC(fpu), IDTVEC(align),
87 IDTVEC(xmm), IDTVEC(lcall_syscall), IDTVEC(int0x80_syscall);
91 start_info_t *xen_start_info;
92 shared_info_t *HYPERVISOR_shared_info;
93 xen_pfn_t *xen_machine_phys = machine_to_phys_mapping;
94 xen_pfn_t *xen_phys_machine;
95 xen_pfn_t *xen_pfn_to_mfn_frame_list[16];
96 xen_pfn_t *xen_pfn_to_mfn_frame_list_list;
97 int preemptable, init_first;
98 extern unsigned int avail_space;
99 int xen_vector_callback_enabled = 0;
100 enum xen_domain_type xen_domain_type = XEN_PV_DOMAIN;
109 CTR0(KTR_SPARE2, "ni_cli disabling interrupts");
110 __asm__("pushl %edx;"
123 __asm__("pushl %edx;"
135 force_evtchn_callback(void)
137 (void)HYPERVISOR_xen_version(0, NULL);
141 * Modify the cmd_line by converting ',' to NULLs so that it is in a format
142 * suitable for the static env vars.
145 xen_setbootenv(char *cmd_line)
149 /* Skip leading spaces */
150 for (; *cmd_line == ' '; cmd_line++);
152 xc_printf("xen_setbootenv(): cmd_line='%s'\n", cmd_line);
154 for (cmd_line_next = cmd_line; strsep(&cmd_line_next, ",") != NULL;);
163 {"boot_askname", RB_ASKNAME},
164 {"boot_single", RB_SINGLE},
165 {"boot_nosync", RB_NOSYNC},
166 {"boot_halt", RB_ASKNAME},
167 {"boot_serial", RB_SERIAL},
168 {"boot_cdrom", RB_CDROM},
169 {"boot_gdb", RB_GDB},
170 {"boot_gdb_pause", RB_RESERVED1},
171 {"boot_verbose", RB_VERBOSE},
172 {"boot_multicons", RB_MULTIPLE},
177 xen_boothowto(char *envp)
181 /* get equivalents from the environment */
182 for (i = 0; howto_names[i].ev != NULL; i++)
183 if (getenv(howto_names[i].ev) != NULL)
184 howto |= howto_names[i].mask;
188 #define XC_PRINTF_BUFSIZE 1024
190 xc_printf(const char *fmt, ...)
194 static char buf[XC_PRINTF_BUFSIZE];
197 retval = vsnprintf(buf, XC_PRINTF_BUFSIZE - 1, fmt, ap);
200 (void)HYPERVISOR_console_write(buf, retval);
204 #define XPQUEUE_SIZE 128
212 /* per-cpu queues and indices */
214 static struct mmu_log xpq_queue_log[XEN_LEGACY_MAX_VCPUS][XPQUEUE_SIZE];
217 static int xpq_idx[XEN_LEGACY_MAX_VCPUS];
218 static mmu_update_t xpq_queue[XEN_LEGACY_MAX_VCPUS][XPQUEUE_SIZE];
220 #define XPQ_QUEUE_LOG xpq_queue_log[vcpu]
221 #define XPQ_QUEUE xpq_queue[vcpu]
222 #define XPQ_IDX xpq_idx[vcpu]
223 #define SET_VCPU() int vcpu = smp_processor_id()
226 static mmu_update_t xpq_queue[XPQUEUE_SIZE];
228 static struct mmu_log xpq_queue_log[XPQUEUE_SIZE];
230 static int xpq_idx = 0;
232 #define XPQ_QUEUE_LOG xpq_queue_log
233 #define XPQ_QUEUE xpq_queue
234 #define XPQ_IDX xpq_idx
238 #define XPQ_IDX_INC atomic_add_int(&XPQ_IDX, 1);
244 int _xpq_idx = XPQ_IDX;
250 xc_printf("xen_dump_queue(): %u entries\n", _xpq_idx);
251 for (i = 0; i < _xpq_idx; i++) {
252 xc_printf(" val: %llx ptr: %llx\n", XPQ_QUEUE[i].val,
260 _xen_flush_queue(void)
263 int _xpq_idx = XPQ_IDX;
267 if (__predict_true(gdtset))
268 CRITICAL_ASSERT(curthread);
272 /* Make sure index is cleared first to avoid double updates. */
273 error = HYPERVISOR_mmu_update((mmu_update_t *)&XPQ_QUEUE,
274 _xpq_idx, NULL, DOMID_SELF);
277 if (__predict_true(gdtset))
278 for (i = _xpq_idx; i > 0;) {
280 CTR6(KTR_PMAP, "mmu:val: %lx ptr: %lx val: %lx "
281 "ptr: %lx val: %lx ptr: %lx",
282 (XPQ_QUEUE[i-1].val & 0xffffffff),
283 (XPQ_QUEUE[i-1].ptr & 0xffffffff),
284 (XPQ_QUEUE[i-2].val & 0xffffffff),
285 (XPQ_QUEUE[i-2].ptr & 0xffffffff),
286 (XPQ_QUEUE[i-3].val & 0xffffffff),
287 (XPQ_QUEUE[i-3].ptr & 0xffffffff));
290 CTR4(KTR_PMAP, "mmu: val: %lx ptr: %lx val: %lx ptr: %lx",
291 (XPQ_QUEUE[i-1].val & 0xffffffff),
292 (XPQ_QUEUE[i-1].ptr & 0xffffffff),
293 (XPQ_QUEUE[i-2].val & 0xffffffff),
294 (XPQ_QUEUE[i-2].ptr & 0xffffffff));
297 CTR2(KTR_PMAP, "mmu: val: %lx ptr: %lx",
298 (XPQ_QUEUE[i-1].val & 0xffffffff),
299 (XPQ_QUEUE[i-1].ptr & 0xffffffff));
304 if (__predict_false(error < 0)) {
305 for (i = 0; i < _xpq_idx; i++)
306 printf("val: %llx ptr: %llx\n",
307 XPQ_QUEUE[i].val, XPQ_QUEUE[i].ptr);
308 panic("Failed to execute MMU updates: %d", error);
314 xen_flush_queue(void)
318 if (__predict_true(gdtset))
320 if (XPQ_IDX != 0) _xen_flush_queue();
321 if (__predict_true(gdtset))
326 xen_increment_idx(void)
331 if (__predict_false(XPQ_IDX == XPQUEUE_SIZE))
336 xen_check_queue(void)
341 KASSERT(XPQ_IDX == 0, ("pending operations XPQ_IDX=%d", XPQ_IDX));
346 xen_invlpg(vm_offset_t va)
349 op.cmd = MMUEXT_INVLPG_ALL;
350 op.arg1.linear_addr = va & ~PAGE_MASK;
351 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
355 xen_load_cr3(u_int val)
361 KASSERT(XPQ_IDX == 0, ("pending operations XPQ_IDX=%d", XPQ_IDX));
363 op.cmd = MMUEXT_NEW_BASEPTR;
364 op.arg1.mfn = xpmap_ptom(val) >> PAGE_SHIFT;
365 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
369 static __inline u_int
374 __asm __volatile("movl 4(%%ebp),%0" : "=r" (data));
385 eflags = _read_eflags();
386 _vcpu = &HYPERVISOR_shared_info->vcpu_info[smp_processor_id()];
387 if (_vcpu->evtchn_upcall_mask)
394 write_eflags(u_int eflags)
398 CTR2(KTR_SPARE2, "%x xen_restore_flags eflags %x", rebp(), eflags);
399 intr = ((eflags & PSL_I) == 0);
400 __restore_flags(intr);
401 _write_eflags(eflags);
407 CTR1(KTR_SPARE2, "%x xen_cli disabling interrupts", rebp());
414 CTR1(KTR_SPARE2, "%x xen_sti enabling interrupts", rebp());
422 return (HYPERVISOR_shared_info->vcpu_info[curcpu].arch.cr2);
426 _xen_machphys_update(vm_paddr_t mfn, vm_paddr_t pfn, char *file, int line)
430 if (__predict_true(gdtset))
432 XPQ_QUEUE[XPQ_IDX].ptr = (mfn << PAGE_SHIFT) | MMU_MACHPHYS_UPDATE;
433 XPQ_QUEUE[XPQ_IDX].val = pfn;
435 XPQ_QUEUE_LOG[XPQ_IDX].file = file;
436 XPQ_QUEUE_LOG[XPQ_IDX].line = line;
439 if (__predict_true(gdtset))
443 extern struct rwlock pvh_global_lock;
446 _xen_queue_pt_update(vm_paddr_t ptr, vm_paddr_t val, char *file, int line)
450 if (__predict_true(gdtset))
451 rw_assert(&pvh_global_lock, RA_WLOCKED);
453 KASSERT((ptr & 7) == 0, ("misaligned update"));
455 if (__predict_true(gdtset))
458 XPQ_QUEUE[XPQ_IDX].ptr = ((uint64_t)ptr) | MMU_NORMAL_PT_UPDATE;
459 XPQ_QUEUE[XPQ_IDX].val = (uint64_t)val;
461 XPQ_QUEUE_LOG[XPQ_IDX].file = file;
462 XPQ_QUEUE_LOG[XPQ_IDX].line = line;
465 if (__predict_true(gdtset))
470 xen_pgdpt_pin(vm_paddr_t ma)
473 op.cmd = MMUEXT_PIN_L3_TABLE;
474 op.arg1.mfn = ma >> PAGE_SHIFT;
476 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
480 xen_pgd_pin(vm_paddr_t ma)
483 op.cmd = MMUEXT_PIN_L2_TABLE;
484 op.arg1.mfn = ma >> PAGE_SHIFT;
486 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
490 xen_pgd_unpin(vm_paddr_t ma)
493 op.cmd = MMUEXT_UNPIN_TABLE;
494 op.arg1.mfn = ma >> PAGE_SHIFT;
496 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
500 xen_pt_pin(vm_paddr_t ma)
503 op.cmd = MMUEXT_PIN_L1_TABLE;
504 op.arg1.mfn = ma >> PAGE_SHIFT;
506 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
510 xen_pt_unpin(vm_paddr_t ma)
513 op.cmd = MMUEXT_UNPIN_TABLE;
514 op.arg1.mfn = ma >> PAGE_SHIFT;
516 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
520 xen_set_ldt(vm_paddr_t ptr, unsigned long len)
523 op.cmd = MMUEXT_SET_LDT;
524 op.arg1.linear_addr = ptr;
525 op.arg2.nr_ents = len;
527 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
530 void xen_tlb_flush(void)
533 op.cmd = MMUEXT_TLB_FLUSH_LOCAL;
535 PANIC_IF(HYPERVISOR_mmuext_op(&op, 1, NULL, DOMID_SELF) < 0);
539 xen_update_descriptor(union descriptor *table, union descriptor *entry)
544 ptp = vtopte((vm_offset_t)table);
545 pa = (*ptp & PG_FRAME) | ((vm_offset_t)table & PAGE_MASK);
546 if (HYPERVISOR_update_descriptor(pa, *(uint64_t *)entry))
547 panic("HYPERVISOR_update_descriptor failed\n");
553 * Bitmap is indexed by page number. If bit is set, the page is part of a
554 * xen_create_contiguous_region() area of memory.
556 unsigned long *contiguous_bitmap;
559 contiguous_bitmap_set(unsigned long first_page, unsigned long nr_pages)
561 unsigned long start_off, end_off, curr_idx, end_idx;
563 curr_idx = first_page / BITS_PER_LONG;
564 start_off = first_page & (BITS_PER_LONG-1);
565 end_idx = (first_page + nr_pages) / BITS_PER_LONG;
566 end_off = (first_page + nr_pages) & (BITS_PER_LONG-1);
568 if (curr_idx == end_idx) {
569 contiguous_bitmap[curr_idx] |=
570 ((1UL<<end_off)-1) & -(1UL<<start_off);
572 contiguous_bitmap[curr_idx] |= -(1UL<<start_off);
573 while ( ++curr_idx < end_idx )
574 contiguous_bitmap[curr_idx] = ~0UL;
575 contiguous_bitmap[curr_idx] |= (1UL<<end_off)-1;
580 contiguous_bitmap_clear(unsigned long first_page, unsigned long nr_pages)
582 unsigned long start_off, end_off, curr_idx, end_idx;
584 curr_idx = first_page / BITS_PER_LONG;
585 start_off = first_page & (BITS_PER_LONG-1);
586 end_idx = (first_page + nr_pages) / BITS_PER_LONG;
587 end_off = (first_page + nr_pages) & (BITS_PER_LONG-1);
589 if (curr_idx == end_idx) {
590 contiguous_bitmap[curr_idx] &=
591 -(1UL<<end_off) | ((1UL<<start_off)-1);
593 contiguous_bitmap[curr_idx] &= (1UL<<start_off)-1;
594 while ( ++curr_idx != end_idx )
595 contiguous_bitmap[curr_idx] = 0;
596 contiguous_bitmap[curr_idx] &= -(1UL<<end_off);
601 /* Ensure multi-page extents are contiguous in machine memory. */
603 xen_create_contiguous_region(vm_page_t pages, int npages)
605 unsigned long mfn, i, flags;
607 struct xen_memory_reservation reservation = {
612 set_xen_guest_handle(reservation.extent_start, &mfn);
616 /* can currently only handle power of two allocation */
617 PANIC_IF(ffs(npages) != fls(npages));
619 /* 0. determine order */
620 order = (ffs(npages) == fls(npages)) ? fls(npages) - 1 : fls(npages);
622 /* 1. give away machine pages. */
623 for (i = 0; i < (1 << order); i++) {
625 pfn = VM_PAGE_TO_PHYS(&pages[i]) >> PAGE_SHIFT;
627 PFNTOMFN(pfn) = INVALID_P2M_ENTRY;
628 PANIC_IF(HYPERVISOR_memory_op(XENMEM_decrease_reservation, &reservation) != 1);
632 /* 2. Get a new contiguous memory extent. */
633 reservation.extent_order = order;
634 /* xenlinux hardcodes this because of aacraid - maybe set to 0 if we're not
635 * running with a broxen driver XXXEN
637 reservation.address_bits = 31;
638 if (HYPERVISOR_memory_op(XENMEM_increase_reservation, &reservation) != 1)
641 /* 3. Map the new extent in place of old pages. */
642 for (i = 0; i < (1 << order); i++) {
644 pfn = VM_PAGE_TO_PHYS(&pages[i]) >> PAGE_SHIFT;
645 xen_machphys_update(mfn+i, pfn);
646 PFNTOMFN(pfn) = mfn+i;
652 contiguous_bitmap_set(VM_PAGE_TO_PHYS(&pages[0]) >> PAGE_SHIFT, 1UL << order);
655 balloon_unlock(flags);
660 reservation.extent_order = 0;
661 reservation.address_bits = 0;
663 for (i = 0; i < (1 << order); i++) {
665 pfn = VM_PAGE_TO_PHYS(&pages[i]) >> PAGE_SHIFT;
666 PANIC_IF(HYPERVISOR_memory_op(
667 XENMEM_increase_reservation, &reservation) != 1);
668 xen_machphys_update(mfn, pfn);
674 balloon_unlock(flags);
680 xen_destroy_contiguous_region(void *addr, int npages)
682 unsigned long mfn, i, flags, order, pfn0;
683 struct xen_memory_reservation reservation = {
688 set_xen_guest_handle(reservation.extent_start, &mfn);
690 pfn0 = vtophys(addr) >> PAGE_SHIFT;
692 scrub_pages(vstart, 1 << order);
694 /* can currently only handle power of two allocation */
695 PANIC_IF(ffs(npages) != fls(npages));
697 /* 0. determine order */
698 order = (ffs(npages) == fls(npages)) ? fls(npages) - 1 : fls(npages);
703 contiguous_bitmap_clear(vtophys(addr) >> PAGE_SHIFT, 1UL << order);
706 /* 1. Zap current PTEs, giving away the underlying pages. */
707 for (i = 0; i < (1 << order); i++) {
709 uint64_t new_val = 0;
710 pfn = vtomach((char *)addr + i*PAGE_SIZE) >> PAGE_SHIFT;
712 PANIC_IF(HYPERVISOR_update_va_mapping((vm_offset_t)((char *)addr + (i * PAGE_SIZE)), new_val, 0));
713 PFNTOMFN(pfn) = INVALID_P2M_ENTRY;
714 PANIC_IF(HYPERVISOR_memory_op(
715 XENMEM_decrease_reservation, &reservation) != 1);
718 /* 2. Map new pages in place of old pages. */
719 for (i = 0; i < (1 << order); i++) {
723 PANIC_IF(HYPERVISOR_memory_op(XENMEM_increase_reservation, &reservation) != 1);
725 new_val = mfn << PAGE_SHIFT;
726 PANIC_IF(HYPERVISOR_update_va_mapping((vm_offset_t)addr + (i * PAGE_SIZE),
727 new_val, PG_KERNEL));
728 xen_machphys_update(mfn, pfn);
734 balloon_unlock(flags);
737 extern vm_offset_t proc0kstack;
738 extern int vm86paddr, vm86phystk;
739 char *bootmem_start, *bootmem_current, *bootmem_end;
741 pteinfo_t *pteinfo_list;
742 void initvalues(start_info_t *startinfo);
744 struct xenstore_domain_interface;
745 extern struct xenstore_domain_interface *xen_store;
750 bootmem_alloc(unsigned int size)
754 retptr = bootmem_current;
755 PANIC_IF(retptr + size > bootmem_end);
756 bootmem_current += size;
762 bootmem_free(void *ptr, unsigned int size)
767 PANIC_IF(tptr != bootmem_current - size ||
768 bootmem_current - size < bootmem_start);
770 bootmem_current -= size;
775 xpmap_mtop2(vm_paddr_t mpa)
777 return ((machine_to_phys_mapping[mpa >> PAGE_SHIFT] << PAGE_SHIFT)
778 ) | (mpa & ~PG_FRAME);
782 xpmap_get_bootpde(vm_paddr_t va)
785 return ((pd_entry_t *)xen_start_info->pt_base)[va >> 22];
789 xpmap_get_vbootpde(vm_paddr_t va)
793 pde = xpmap_get_bootpde(va);
794 if ((pde & PG_V) == 0)
795 return (pde & ~PG_FRAME);
796 return (pde & ~PG_FRAME) |
797 (xpmap_mtop2(pde & PG_FRAME) + KERNBASE);
801 xpmap_get_bootptep(vm_paddr_t va)
805 pde = xpmap_get_vbootpde(va);
806 if ((pde & PG_V) == 0)
808 #define PT_MASK 0x003ff000 /* page table address bits */
809 return &(((pt_entry_t *)(pde & PG_FRAME))[(va & PT_MASK) >> PAGE_SHIFT]);
813 xpmap_get_bootpte(vm_paddr_t va)
816 return xpmap_get_bootptep(va)[0];
823 shift_phys_machine(unsigned long *phys_machine, int nr_pages)
826 unsigned long *tmp_page, *current_page, *next_page;
829 tmp_page = bootmem_alloc(PAGE_SIZE);
830 current_page = phys_machine + nr_pages - (PAGE_SIZE/sizeof(unsigned long));
831 next_page = current_page - (PAGE_SIZE/sizeof(unsigned long));
832 bcopy(phys_machine, tmp_page, PAGE_SIZE);
834 while (current_page > phys_machine) {
836 bcopy(next_page, tmp_page, PAGE_SIZE);
837 /* shift down page */
838 bcopy(current_page, next_page, PAGE_SIZE);
840 bcopy(tmp_page, current_page, PAGE_SIZE);
842 current_page -= (PAGE_SIZE/sizeof(unsigned long));
843 next_page -= (PAGE_SIZE/sizeof(unsigned long));
845 bootmem_free(tmp_page, PAGE_SIZE);
847 for (i = 0; i < nr_pages; i++) {
848 xen_machphys_update(phys_machine[i], i);
850 memset(phys_machine, INVALID_P2M_ENTRY, PAGE_SIZE);
853 #endif /* ADD_ISA_HOLE */
856 * Build a directory of the pages that make up our Physical to Machine
857 * mapping table. The Xen suspend/restore code uses this to find our
861 init_frame_list_list(void *arg)
863 unsigned long nr_pages = xen_start_info->nr_pages;
864 #define FPP (PAGE_SIZE/sizeof(xen_pfn_t))
867 xen_pfn_to_mfn_frame_list_list = malloc(PAGE_SIZE, M_DEVBUF, M_WAITOK);
868 for (i = 0, j = 0, k = -1; i < nr_pages;
870 if ((j & (FPP - 1)) == 0) {
872 xen_pfn_to_mfn_frame_list[k] =
873 malloc(PAGE_SIZE, M_DEVBUF, M_WAITOK);
874 xen_pfn_to_mfn_frame_list_list[k] =
875 VTOMFN(xen_pfn_to_mfn_frame_list[k]);
878 xen_pfn_to_mfn_frame_list[k][j] =
879 VTOMFN(&xen_phys_machine[i]);
882 HYPERVISOR_shared_info->arch.max_pfn = nr_pages;
883 HYPERVISOR_shared_info->arch.pfn_to_mfn_frame_list_list
884 = VTOMFN(xen_pfn_to_mfn_frame_list_list);
886 SYSINIT(init_fll, SI_SUB_DEVFS, SI_ORDER_ANY, init_frame_list_list, NULL);
888 extern unsigned long physfree;
893 extern uint32_t kernbase;
896 initvalues(start_info_t *startinfo)
898 vm_offset_t cur_space, cur_space_pt;
899 struct physdev_set_iopl set_iopl;
901 int l3_pages, l2_pages, l1_pages, offset;
902 vm_paddr_t console_page_ma, xen_store_ma;
906 vm_paddr_t IdlePDPTma, IdlePDPTnewma;
907 vm_paddr_t IdlePTDnewma[4];
908 pd_entry_t *IdlePDPTnew, *IdlePTDnew;
909 vm_paddr_t IdlePTDma[4];
911 vm_paddr_t IdlePTDma[1];
918 max((startinfo->nr_pages >> NPGPTD_SHIFT), nkpt),
919 NPGPTD*NPDEPG - KPTDI),
920 (HYPERVISOR_VIRT_START - KERNBASE) >> PDRSHIFT);
922 HYPERVISOR_vm_assist(VMASST_CMD_enable, VMASST_TYPE_4gb_segments);
925 * need to install handler
927 HYPERVISOR_vm_assist(VMASST_CMD_enable, VMASST_TYPE_4gb_segments_notify);
929 xen_start_info = startinfo;
930 xen_phys_machine = (xen_pfn_t *)startinfo->mfn_list;
932 IdlePTD = (pd_entry_t *)((uint8_t *)startinfo->pt_base + PAGE_SIZE);
938 IdlePDPT = (pd_entry_t *)startinfo->pt_base;
939 IdlePDPTma = VTOM(startinfo->pt_base);
940 for (i = (KERNBASE >> 30);
941 (i < 4) && (IdlePDPT[i] != 0); i++)
944 * Note that only one page directory has been allocated at this point.
947 for (i = 0; i < l2_pages; i++)
948 IdlePTDma[i] = VTOM(IdlePTD + i*PAGE_SIZE);
950 l2_pages = (l2_pages == 0) ? 1 : l2_pages;
955 for (i = (((KERNBASE>>18) & PAGE_MASK)>>PAGE_SHIFT);
956 (i<l2_pages*NPDEPG) && (i<(VM_MAX_KERNEL_ADDRESS>>PDRSHIFT)); i++) {
963 /* number of pages allocated after the pts + 1*/;
964 cur_space = xen_start_info->pt_base +
965 (l3_pages + l2_pages + l1_pages + 1)*PAGE_SIZE;
967 xc_printf("initvalues(): wooh - availmem=%x,%x\n", avail_space,
970 xc_printf("KERNBASE=%x,pt_base=%x, VTOPFN(base)=%x, nr_pt_frames=%x\n",
971 KERNBASE,xen_start_info->pt_base, VTOPFN(xen_start_info->pt_base),
972 xen_start_info->nr_pt_frames);
973 xendebug_flags = 0; /* 0xffffffff; */
976 shift_phys_machine(xen_phys_machine, xen_start_info->nr_pages);
978 XENPRINTF("IdlePTD %p\n", IdlePTD);
979 XENPRINTF("nr_pages: %ld shared_info: 0x%lx flags: 0x%lx pt_base: 0x%lx "
980 "mod_start: 0x%lx mod_len: 0x%lx\n",
981 xen_start_info->nr_pages, xen_start_info->shared_info,
982 xen_start_info->flags, xen_start_info->pt_base,
983 xen_start_info->mod_start, xen_start_info->mod_len);
986 IdlePDPTnew = (pd_entry_t *)cur_space; cur_space += PAGE_SIZE;
987 bzero(IdlePDPTnew, PAGE_SIZE);
989 IdlePDPTnewma = VTOM(IdlePDPTnew);
990 IdlePTDnew = (pd_entry_t *)cur_space; cur_space += 4*PAGE_SIZE;
991 bzero(IdlePTDnew, 4*PAGE_SIZE);
993 for (i = 0; i < 4; i++)
994 IdlePTDnewma[i] = VTOM((uint8_t *)IdlePTDnew + i*PAGE_SIZE);
998 * Copy the 4 machine addresses of the new PTDs in to the PDPT
1001 for (i = 0; i < 4; i++)
1002 IdlePDPTnew[i] = IdlePTDnewma[i] | PG_V;
1007 * re-map the new PDPT read-only
1009 PT_SET_MA(IdlePDPTnew, IdlePDPTnewma | PG_V);
1012 * Unpin the current PDPT
1014 xen_pt_unpin(IdlePDPTma);
1018 /* Map proc0's KSTACK */
1019 proc0kstack = cur_space; cur_space += (KSTACK_PAGES * PAGE_SIZE);
1020 xc_printf("proc0kstack=%u\n", proc0kstack);
1022 /* vm86/bios stack */
1023 cur_space += PAGE_SIZE;
1025 /* Map space for the vm86 region */
1026 vm86paddr = (vm_offset_t)cur_space;
1027 cur_space += (PAGE_SIZE * 3);
1029 /* allocate 4 pages for bootmem allocator */
1030 bootmem_start = bootmem_current = (char *)cur_space;
1031 cur_space += (4 * PAGE_SIZE);
1032 bootmem_end = (char *)cur_space;
1034 /* allocate pages for gdt */
1035 gdt = (union descriptor *)cur_space;
1036 cur_space += PAGE_SIZE*ncpus;
1038 /* allocate page for ldt */
1039 ldt = (union descriptor *)cur_space; cur_space += PAGE_SIZE;
1040 cur_space += PAGE_SIZE;
1042 /* unmap remaining pages from initial chunk
1045 for (tmpva = cur_space; tmpva < (((uint32_t)&kernbase) + (l1_pages<<PDRSHIFT));
1046 tmpva += PAGE_SIZE) {
1047 bzero((char *)tmpva, PAGE_SIZE);
1048 PT_SET_MA(tmpva, (vm_paddr_t)0);
1053 memcpy(((uint8_t *)IdlePTDnew) + ((unsigned int)(KERNBASE >> 18)),
1054 ((uint8_t *)IdlePTD) + ((KERNBASE >> 18) & PAGE_MASK),
1055 l1_pages*sizeof(pt_entry_t));
1057 for (i = 0; i < 4; i++) {
1058 PT_SET_MA((uint8_t *)IdlePTDnew + i*PAGE_SIZE,
1059 IdlePTDnewma[i] | PG_V);
1061 xen_load_cr3(VTOP(IdlePDPTnew));
1062 xen_pgdpt_pin(VTOM(IdlePDPTnew));
1064 /* allocate remainder of nkpt pages */
1065 cur_space_pt = cur_space;
1066 for (offset = (KERNBASE >> PDRSHIFT), i = l1_pages; i < nkpt;
1067 i++, cur_space += PAGE_SIZE) {
1068 pdir = (offset + i) / NPDEPG;
1069 curoffset = ((offset + i) % NPDEPG);
1070 if (((offset + i) << PDRSHIFT) == VM_MAX_KERNEL_ADDRESS)
1074 * make sure that all the initial page table pages
1077 PT_SET_MA(cur_space, VTOM(cur_space) | PG_V | PG_RW);
1078 bzero((char *)cur_space, PAGE_SIZE);
1079 PT_SET_MA(cur_space, (vm_paddr_t)0);
1080 xen_pt_pin(VTOM(cur_space));
1081 xen_queue_pt_update((vm_paddr_t)(IdlePTDnewma[pdir] +
1082 curoffset*sizeof(vm_paddr_t)),
1083 VTOM(cur_space) | PG_KERNEL);
1087 for (i = 0; i < 4; i++) {
1088 pdir = (PTDPTDI + i) / NPDEPG;
1089 curoffset = (PTDPTDI + i) % NPDEPG;
1091 xen_queue_pt_update((vm_paddr_t)(IdlePTDnewma[pdir] +
1092 curoffset*sizeof(vm_paddr_t)),
1093 IdlePTDnewma[i] | PG_V);
1098 IdlePTD = IdlePTDnew;
1099 IdlePDPT = IdlePDPTnew;
1100 IdlePDPTma = IdlePDPTnewma;
1102 HYPERVISOR_shared_info = (shared_info_t *)cur_space;
1103 cur_space += PAGE_SIZE;
1105 xen_store = (struct xenstore_domain_interface *)cur_space;
1106 cur_space += PAGE_SIZE;
1108 console_page = (char *)cur_space;
1109 cur_space += PAGE_SIZE;
1112 * shared_info is an unsigned long so this will randomly break if
1113 * it is allocated above 4GB - I guess people are used to that
1114 * sort of thing with Xen ... sigh
1116 shinfo = xen_start_info->shared_info;
1117 PT_SET_MA(HYPERVISOR_shared_info, shinfo | PG_KERNEL);
1121 xen_store_ma = (((vm_paddr_t)xen_start_info->store_mfn) << PAGE_SHIFT);
1122 PT_SET_MA(xen_store, xen_store_ma | PG_KERNEL);
1123 console_page_ma = (((vm_paddr_t)xen_start_info->console.domU.mfn) << PAGE_SHIFT);
1124 PT_SET_MA(console_page, console_page_ma | PG_KERNEL);
1129 PANIC_IF(HYPERVISOR_physdev_op(PHYSDEVOP_SET_IOPL, &set_iopl));
1132 /* add page table for KERNBASE */
1133 xen_queue_pt_update(IdlePTDma + KPTDI*sizeof(vm_paddr_t),
1134 VTOM(cur_space) | PG_KERNEL);
1137 xen_queue_pt_update(pdir_shadow_ma[3] + KPTDI*sizeof(vm_paddr_t),
1138 VTOM(cur_space) | PG_V | PG_A);
1140 xen_queue_pt_update(pdir_shadow_ma + KPTDI*sizeof(vm_paddr_t),
1141 VTOM(cur_space) | PG_V | PG_A);
1144 cur_space += PAGE_SIZE;
1148 if (xen_start_info->flags & SIF_INITDOMAIN) {
1149 /* Map first megabyte */
1150 for (i = 0; i < (256 << PAGE_SHIFT); i += PAGE_SIZE)
1151 PT_SET_MA(KERNBASE + i, i | PG_KERNEL | PG_NC_PCD);
1156 * re-map kernel text read-only
1159 for (i = (((vm_offset_t)&btext) & ~PAGE_MASK);
1160 i < (((vm_offset_t)&etext) & ~PAGE_MASK); i += PAGE_SIZE)
1161 PT_SET_MA(i, VTOM(i) | PG_V | PG_A);
1164 physfree = VTOP(cur_space);
1165 init_first = physfree >> PAGE_SHIFT;
1166 IdlePTD = (pd_entry_t *)VTOP(IdlePTD);
1167 IdlePDPT = (pd_entry_t *)VTOP(IdlePDPT);
1168 setup_xen_features();
1169 xc_printf("#8, proc0kstack=%u\n", proc0kstack);
1173 trap_info_t trap_table[] = {
1174 { 0, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(div)},
1175 { 1, 0|4, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(dbg)},
1176 { 3, 3|4, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(bpt)},
1177 { 4, 3, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(ofl)},
1178 /* This is UPL on Linux and KPL on BSD */
1179 { 5, 3, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(bnd)},
1180 { 6, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(ill)},
1181 { 7, 0|4, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(dna)},
1183 * { 8, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(XXX)},
1184 * no handler for double fault
1186 { 9, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(fpusegm)},
1187 {10, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(tss)},
1188 {11, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(missing)},
1189 {12, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(stk)},
1190 {13, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(prot)},
1191 {14, 0|4, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(page)},
1192 {15, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(rsvd)},
1193 {16, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(fpu)},
1194 {17, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(align)},
1195 {18, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(mchk)},
1196 {19, 0, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(xmm)},
1197 {0x80, 3, GSEL(GCODE_SEL, SEL_KPL), (unsigned long) &IDTVEC(int0x80_syscall)},
1201 /* Perform a multicall and check that individual calls succeeded. */
1203 HYPERVISOR_multicall(struct multicall_entry * call_list, int nr_calls)
1208 /* Perform the multicall. */
1209 PANIC_IF(_HYPERVISOR_multicall(call_list, nr_calls));
1211 /* Check the results of individual hypercalls. */
1212 for (i = 0; i < nr_calls; i++)
1213 if (__predict_false(call_list[i].result < 0))
1215 if (__predict_false(ret > 0))
1216 panic("%d multicall(s) failed: cpu %d\n",
1217 ret, smp_processor_id());
1219 /* If we didn't panic already, everything succeeded. */
1223 /********** CODE WORTH KEEPING ABOVE HERE *****************/
1225 void xen_failsafe_handler(void);
1228 xen_failsafe_handler(void)
1231 panic("xen_failsafe_handler called!\n");
1234 void xen_handle_thread_switch(struct pcb *pcb);
1236 /* This is called by cpu_switch() when switching threads. */
1237 /* The pcb arg refers to the process control block of the */
1238 /* next thread which is to run */
1240 xen_handle_thread_switch(struct pcb *pcb)
1242 uint32_t *a = (uint32_t *)&PCPU_GET(fsgs_gdt)[0];
1243 uint32_t *b = (uint32_t *)&pcb->pcb_fsd;
1244 multicall_entry_t mcl[3];
1247 /* Notify Xen of task switch */
1248 mcl[i].op = __HYPERVISOR_stack_switch;
1249 mcl[i].args[0] = GSEL(GDATA_SEL, SEL_KPL);
1250 mcl[i++].args[1] = (unsigned long)pcb;
1252 /* Check for update of fsd */
1253 if (*a != *b || *(a+1) != *(b+1)) {
1254 mcl[i].op = __HYPERVISOR_update_descriptor;
1255 *(uint64_t *)&mcl[i].args[0] = vtomach((vm_offset_t)a);
1256 *(uint64_t *)&mcl[i++].args[2] = *(uint64_t *)b;
1262 /* Check for update of gsd */
1263 if (*a != *b || *(a+1) != *(b+1)) {
1264 mcl[i].op = __HYPERVISOR_update_descriptor;
1265 *(uint64_t *)&mcl[i].args[0] = vtomach((vm_offset_t)a);
1266 *(uint64_t *)&mcl[i++].args[2] = *(uint64_t *)b;
1269 (void)HYPERVISOR_multicall(mcl, i);