2 * Copyright (c) 2004 Marcel Moolenaar
3 * Copyright (c) 2001 Doug Rabson
4 * Copyright (c) 2016 The FreeBSD Foundation
7 * Portions of this software were developed by Konstantin Belousov
8 * under sponsorship from the FreeBSD Foundation.
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
13 * 1. Redistributions of source code must retain the above copyright
14 * notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 * notice, this list of conditions and the following disclaimer in the
17 * documentation and/or other materials provided with the distribution.
19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
35 #include <sys/param.h>
37 #include <sys/kernel.h>
38 #include <sys/linker.h>
40 #include <sys/module.h>
41 #include <sys/mutex.h>
42 #include <sys/clock.h>
44 #include <sys/rwlock.h>
45 #include <sys/sched.h>
46 #include <sys/sysctl.h>
47 #include <sys/systm.h>
48 #include <sys/vmmeter.h>
50 #include <machine/fpu.h>
51 #include <machine/efi.h>
52 #include <machine/metadata.h>
53 #include <machine/md_var.h>
54 #include <machine/smp.h>
55 #include <machine/vmparam.h>
58 #include <vm/vm_map.h>
59 #include <vm/vm_object.h>
60 #include <vm/vm_page.h>
61 #include <vm/vm_pager.h>
63 static pml4_entry_t *efi_pml4;
64 static vm_object_t obj_1t1_pt;
65 static vm_page_t efi_pml4_page;
66 static vm_pindex_t efi_1t1_idx;
69 efi_destroy_1t1_map(void)
73 if (obj_1t1_pt != NULL) {
74 VM_OBJECT_RLOCK(obj_1t1_pt);
75 TAILQ_FOREACH(m, &obj_1t1_pt->memq, listq)
77 vm_wire_sub(obj_1t1_pt->resident_page_count);
78 VM_OBJECT_RUNLOCK(obj_1t1_pt);
79 vm_object_deallocate(obj_1t1_pt);
91 return (vm_page_grab(obj_1t1_pt, efi_1t1_idx++, VM_ALLOC_NOBUSY |
92 VM_ALLOC_WIRED | VM_ALLOC_ZERO));
96 efi_1t1_pte(vm_offset_t va)
103 vm_pindex_t pml4_idx, pdp_idx, pd_idx;
106 pml4_idx = pmap_pml4e_index(va);
107 pml4e = &efi_pml4[pml4_idx];
110 mphys = VM_PAGE_TO_PHYS(m);
111 *pml4e = mphys | X86_PG_RW | X86_PG_V;
113 mphys = *pml4e & ~PAGE_MASK;
116 pdpe = (pdp_entry_t *)PHYS_TO_DMAP(mphys);
117 pdp_idx = pmap_pdpe_index(va);
121 mphys = VM_PAGE_TO_PHYS(m);
122 *pdpe = mphys | X86_PG_RW | X86_PG_V;
124 mphys = *pdpe & ~PAGE_MASK;
127 pde = (pd_entry_t *)PHYS_TO_DMAP(mphys);
128 pd_idx = pmap_pde_index(va);
132 mphys = VM_PAGE_TO_PHYS(m);
133 *pde = mphys | X86_PG_RW | X86_PG_V;
135 mphys = *pde & ~PAGE_MASK;
138 pte = (pt_entry_t *)PHYS_TO_DMAP(mphys);
139 pte += pmap_pte_index(va);
140 KASSERT(*pte == 0, ("va %#jx *pt %#jx", va, *pte));
146 efi_create_1t1_map(struct efi_md *map, int ndesc, int descsz)
154 obj_1t1_pt = vm_pager_allocate(OBJT_PHYS, NULL, ptoa(1 +
155 NPML4EPG + NPML4EPG * NPDPEPG + NPML4EPG * NPDPEPG * NPDEPG),
156 VM_PROT_ALL, 0, NULL);
158 VM_OBJECT_WLOCK(obj_1t1_pt);
159 efi_pml4_page = efi_1t1_page();
160 VM_OBJECT_WUNLOCK(obj_1t1_pt);
161 efi_pml4 = (pml4_entry_t *)PHYS_TO_DMAP(VM_PAGE_TO_PHYS(efi_pml4_page));
162 pmap_pinit_pml4(efi_pml4_page);
164 for (i = 0, p = map; i < ndesc; i++, p = efi_next_descriptor(p,
166 if ((p->md_attr & EFI_MD_ATTR_RT) == 0)
168 if (p->md_virt != NULL && (uint64_t)p->md_virt != p->md_phys) {
170 printf("EFI Runtime entry %d is mapped\n", i);
173 if ((p->md_phys & EFI_PAGE_MASK) != 0) {
175 printf("EFI Runtime entry %d is not aligned\n",
179 if (p->md_phys + p->md_pages * EFI_PAGE_SIZE < p->md_phys ||
180 p->md_phys + p->md_pages * EFI_PAGE_SIZE >=
181 VM_MAXUSER_ADDRESS) {
182 printf("EFI Runtime entry %d is not in mappable for RT:"
183 "base %#016jx %#jx pages\n",
184 i, (uintmax_t)p->md_phys,
185 (uintmax_t)p->md_pages);
188 if ((p->md_attr & EFI_MD_ATTR_WB) != 0)
189 mode = VM_MEMATTR_WRITE_BACK;
190 else if ((p->md_attr & EFI_MD_ATTR_WT) != 0)
191 mode = VM_MEMATTR_WRITE_THROUGH;
192 else if ((p->md_attr & EFI_MD_ATTR_WC) != 0)
193 mode = VM_MEMATTR_WRITE_COMBINING;
194 else if ((p->md_attr & EFI_MD_ATTR_WP) != 0)
195 mode = VM_MEMATTR_WRITE_PROTECTED;
196 else if ((p->md_attr & EFI_MD_ATTR_UC) != 0)
197 mode = VM_MEMATTR_UNCACHEABLE;
200 printf("EFI Runtime entry %d mapping "
201 "attributes unsupported\n", i);
202 mode = VM_MEMATTR_UNCACHEABLE;
204 bits = pmap_cache_bits(kernel_pmap, mode, FALSE) | X86_PG_RW |
206 VM_OBJECT_WLOCK(obj_1t1_pt);
207 for (va = p->md_phys, idx = 0; idx < p->md_pages; idx++,
209 pte = efi_1t1_pte(va);
210 pte_store(pte, va | bits);
212 VM_OBJECT_WUNLOCK(obj_1t1_pt);
218 efi_destroy_1t1_map();
223 * Create an environment for the EFI runtime code call. The most
224 * important part is creating the required 1:1 physical->virtual
225 * mappings for the runtime segments. To do that, we manually create
226 * page table which unmap userspace but gives correct kernel mapping.
227 * The 1:1 mappings for runtime segments usually occupy low 4G of the
228 * physical address map.
230 * The 1:1 mappings were chosen over the SetVirtualAddressMap() EFI RT
231 * service, because there are some BIOSes which fail to correctly
232 * relocate itself on the call, requiring both 1:1 and virtual
233 * mapping. As result, we must provide 1:1 mapping anyway, so no
234 * reason to bother with the virtual map, and no need to add a
235 * complexity into loader.
237 * The fpu_kern_enter() call allows firmware to use FPU, as mandated
238 * by the specification. In particular, CR0.TS bit is cleared. Also
239 * it enters critical section, giving us neccessary protection against
242 * There is no need to disable interrupts around the change of %cr3,
243 * the kernel mappings are correct, while we only grabbed the
244 * userspace portion of VA. Interrupts handlers must not access
245 * userspace. Having interrupts enabled fixes the issue with
246 * firmware/SMM long operation, which would negatively affect IPIs,
247 * esp. TLB shootdown requests.
254 curpmap = PCPU_GET(curpmap);
255 PMAP_LOCK_ASSERT(curpmap, MA_OWNED);
258 * IPI TLB shootdown handler invltlb_pcid_handler() reloads
259 * %cr3 from the curpmap->pm_cr3, which would disable runtime
260 * segments mappings. Block the handler's action by setting
261 * curpmap to impossible value. See also comment in
262 * pmap.c:pmap_activate_sw().
264 if (pmap_pcid_enabled && !invpcid_works)
265 PCPU_SET(curpmap, NULL);
267 load_cr3(VM_PAGE_TO_PHYS(efi_pml4_page) | (pmap_pcid_enabled ?
268 curpmap->pm_pcids[PCPU_GET(cpuid)].pm_pcid : 0));
270 * If PCID is enabled, the clear CR3_PCID_SAVE bit in the loaded %cr3
271 * causes TLB invalidation.
273 if (!pmap_pcid_enabled)
283 curpmap = &curproc->p_vmspace->vm_pmap;
284 if (pmap_pcid_enabled && !invpcid_works)
285 PCPU_SET(curpmap, curpmap);
286 load_cr3(curpmap->pm_cr3 | (pmap_pcid_enabled ?
287 curpmap->pm_pcids[PCPU_GET(cpuid)].pm_pcid : 0));
288 if (!pmap_pcid_enabled)
292 /* XXX debug stuff */
294 efi_time_sysctl_handler(SYSCTL_HANDLER_ARGS)
300 error = sysctl_handle_int(oidp, &val, 0, req);
301 if (error != 0 || req->newptr == NULL)
303 error = efi_get_time(&tm);
305 uprintf("EFI reports: Year %d Month %d Day %d Hour %d Min %d "
306 "Sec %d\n", tm.tm_year, tm.tm_mon, tm.tm_mday, tm.tm_hour,
307 tm.tm_min, tm.tm_sec);
312 SYSCTL_PROC(_debug, OID_AUTO, efi_time, CTLTYPE_INT | CTLFLAG_RW, NULL, 0,
313 efi_time_sysctl_handler, "I", "");