1/*-
2 * Copyright (c) 2013 The FreeBSD Foundation
3 * All rights reserved.
4 *
5 * This software was developed by Benno Rice under sponsorship from
6 * the FreeBSD Foundation.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
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 *
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
17 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
20 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
26 * SUCH DAMAGE.
27 */
28
29#include <sys/cdefs.h>
30__FBSDID("$FreeBSD$");
31
32#include <sys/param.h>
33
34#include <stand.h>
35#include <bootstrap.h>
36
37#include <efi.h>
38#include <efilib.h>
39
40#include "loader_efi.h"
41
42#if defined(__i386__) || defined(__amd64__)
43#include <machine/cpufunc.h>
44#include <machine/specialreg.h>
45#include <machine/vmparam.h>
46
47/*
48 * The code is excerpted from sys/x86/x86/identcpu.c: identify_cpu(),
49 * identify_hypervisor(), and dev/hyperv/vmbus/hyperv.c: hyperv_identify().
50 */
51#define CPUID_LEAF_HV_MAXLEAF		0x40000000
52#define CPUID_LEAF_HV_INTERFACE		0x40000001
53#define CPUID_LEAF_HV_FEATURES		0x40000003
54#define CPUID_LEAF_HV_LIMITS		0x40000005
55#define CPUID_HV_IFACE_HYPERV		0x31237648	/* HV#1 */
56#define CPUID_HV_MSR_TIME_REFCNT	0x0002	/* MSR_HV_TIME_REF_COUNT */
57#define CPUID_HV_MSR_HYPERCALL		0x0020
58
59static int
60running_on_hyperv(void)
61{
62	char hv_vendor[16];
63	uint32_t regs[4];
64
65	do_cpuid(1, regs);
66	if ((regs[2] & CPUID2_HV) == 0)
67		return (0);
68
69	do_cpuid(CPUID_LEAF_HV_MAXLEAF, regs);
70	if (regs[0] < CPUID_LEAF_HV_LIMITS)
71		return (0);
72
73	((uint32_t *)&hv_vendor)[0] = regs[1];
74	((uint32_t *)&hv_vendor)[1] = regs[2];
75	((uint32_t *)&hv_vendor)[2] = regs[3];
76	hv_vendor[12] = '\0';
77	if (strcmp(hv_vendor, "Microsoft Hv") != 0)
78		return (0);
79
80	do_cpuid(CPUID_LEAF_HV_INTERFACE, regs);
81	if (regs[0] != CPUID_HV_IFACE_HYPERV)
82		return (0);
83
84	do_cpuid(CPUID_LEAF_HV_FEATURES, regs);
85	if ((regs[0] & CPUID_HV_MSR_HYPERCALL) == 0)
86		return (0);
87	if ((regs[0] & CPUID_HV_MSR_TIME_REFCNT) == 0)
88		return (0);
89
90	return (1);
91}
92
93static void
94efi_verify_staging_size(unsigned long *nr_pages)
95{
96	UINTN sz;
97	EFI_MEMORY_DESCRIPTOR *map = NULL, *p;
98	EFI_PHYSICAL_ADDRESS start, end;
99	UINTN key, dsz;
100	UINT32 dver;
101	EFI_STATUS status;
102	int i, ndesc;
103	unsigned long available_pages = 0;
104
105	sz = 0;
106
107	for (;;) {
108		status = BS->GetMemoryMap(&sz, map, &key, &dsz, &dver);
109		if (!EFI_ERROR(status))
110			break;
111
112		if (status != EFI_BUFFER_TOO_SMALL) {
113			printf("Can't read memory map: %lu\n",
114			    EFI_ERROR_CODE(status));
115			goto out;
116		}
117
118		free(map);
119
120		/* Allocate 10 descriptors more than the size reported,
121		 * to allow for any fragmentation caused by calling
122		 * malloc */
123		map = malloc(sz + (10 * dsz));
124		if (map == NULL) {
125			printf("Unable to allocate memory\n");
126			goto out;
127		}
128	}
129
130	ndesc = sz / dsz;
131	for (i = 0, p = map; i < ndesc;
132	     i++, p = NextMemoryDescriptor(p, dsz)) {
133		start = p->PhysicalStart;
134		end = start + p->NumberOfPages * EFI_PAGE_SIZE;
135
136		if (KERNLOAD < start || KERNLOAD >= end)
137			continue;
138
139		available_pages = p->NumberOfPages -
140			((KERNLOAD - start) >> EFI_PAGE_SHIFT);
141		break;
142	}
143
144	if (available_pages == 0) {
145		printf("Can't find valid memory map for staging area!\n");
146		goto out;
147	}
148
149	i++;
150	p = NextMemoryDescriptor(p, dsz);
151
152	for ( ; i < ndesc;
153	     i++, p = NextMemoryDescriptor(p, dsz)) {
154		if (p->Type != EfiConventionalMemory &&
155		    p->Type != EfiLoaderData)
156			break;
157
158		if (p->PhysicalStart != end)
159			break;
160
161		end = p->PhysicalStart + p->NumberOfPages * EFI_PAGE_SIZE;
162
163		available_pages += p->NumberOfPages;
164	}
165
166	if (*nr_pages > available_pages) {
167		printf("Staging area's size is reduced: %ld -> %ld!\n",
168		    *nr_pages, available_pages);
169		*nr_pages = available_pages;
170	}
171out:
172	free(map);
173}
174#endif /* __i386__ || __amd64__ */
175
176#ifndef EFI_STAGING_SIZE
177#if defined(__arm__)
178#define	EFI_STAGING_SIZE	32
179#else
180#define	EFI_STAGING_SIZE	64
181#endif
182#endif
183
184EFI_PHYSICAL_ADDRESS	staging, staging_end, staging_base;
185int			stage_offset_set = 0;
186ssize_t			stage_offset;
187
188int
189efi_copy_init(void)
190{
191	EFI_STATUS	status;
192
193	unsigned long nr_pages;
194
195	nr_pages = EFI_SIZE_TO_PAGES((EFI_STAGING_SIZE) * 1024 * 1024);
196
197#if defined(__i386__) || defined(__amd64__)
198	/*
199	 * We'll decrease nr_pages, if it's too big. Currently we only
200	 * apply this to FreeBSD VM running on Hyper-V. Why? Please see
201	 * https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=211746#c28
202	 */
203	if (running_on_hyperv())
204		efi_verify_staging_size(&nr_pages);
205
206	/*
207	 * The staging area must reside in the the first 1GB physical
208	 * memory: see elf64_exec() in
209	 * boot/efi/loader/arch/amd64/elf64_freebsd.c.
210	 */
211	staging = 1024*1024*1024;
212	status = BS->AllocatePages(AllocateMaxAddress, EfiLoaderData,
213	    nr_pages, &staging);
214#else
215	status = BS->AllocatePages(AllocateAnyPages, EfiLoaderData,
216	    nr_pages, &staging);
217#endif
218	if (EFI_ERROR(status)) {
219		printf("failed to allocate staging area: %lu\n",
220		    EFI_ERROR_CODE(status));
221		return (status);
222	}
223	staging_base = staging;
224	staging_end = staging + nr_pages * EFI_PAGE_SIZE;
225
226#if defined(__aarch64__) || defined(__arm__) || defined(__riscv)
227	/*
228	 * Round the kernel load address to a 2MiB value. This is needed
229	 * because the kernel builds a page table based on where it has
230	 * been loaded in physical address space. As the kernel will use
231	 * either a 1MiB or 2MiB page for this we need to make sure it
232	 * is correctly aligned for both cases.
233	 */
234	staging = roundup2(staging, 2 * 1024 * 1024);
235#endif
236
237	return (0);
238}
239
240static bool
241efi_check_space(vm_offset_t end)
242{
243	EFI_PHYSICAL_ADDRESS addr;
244	EFI_STATUS status;
245	unsigned long nr_pages;
246
247	/* There is already enough space */
248	if (end <= staging_end)
249		return (true);
250
251	end = roundup2(end, EFI_PAGE_SIZE);
252	nr_pages = EFI_SIZE_TO_PAGES(end - staging_end);
253
254#if defined(__i386__) || defined(__amd64__)
255	/* X86 needs all memory to be allocated under the 1G boundary */
256	if (end > 1024*1024*1024)
257		goto before_staging;
258#endif
259
260	/* Try to allocate more space after the previous allocation */
261	addr = staging_end;
262	status = BS->AllocatePages(AllocateAddress, EfiLoaderData, nr_pages,
263	    &addr);
264	if (!EFI_ERROR(status)) {
265		staging_end = staging_end + nr_pages * EFI_PAGE_SIZE;
266		return (true);
267	}
268
269before_staging:
270	/* Try allocating space before the previous allocation */
271	if (staging < nr_pages * EFI_PAGE_SIZE) {
272		printf("Not enough space before allocation\n");
273		return (false);
274	}
275	addr = staging - nr_pages * EFI_PAGE_SIZE;
276#if defined(__aarch64__) || defined(__arm__) || defined(__riscv)
277	/* See efi_copy_init for why this is needed */
278	addr = rounddown2(addr, 2 * 1024 * 1024);
279#endif
280	nr_pages = EFI_SIZE_TO_PAGES(staging_base - addr);
281	status = BS->AllocatePages(AllocateAddress, EfiLoaderData, nr_pages,
282	    &addr);
283	if (!EFI_ERROR(status)) {
284		/*
285		 * Move the old allocation and update the state so
286		 * translation still works.
287		 */
288		staging_base = addr;
289		memmove((void *)(uintptr_t)staging_base,
290		    (void *)(uintptr_t)staging, staging_end - staging);
291		stage_offset -= (staging - staging_base);
292		staging = staging_base;
293		return (true);
294	}
295
296	printf("efi_check_space: Unable to expand staging area\n");
297	return (false);
298}
299
300void *
301efi_translate(vm_offset_t ptr)
302{
303
304	return ((void *)(ptr + stage_offset));
305}
306
307ssize_t
308efi_copyin(const void *src, vm_offset_t dest, const size_t len)
309{
310
311	if (!stage_offset_set) {
312		stage_offset = (vm_offset_t)staging - dest;
313		stage_offset_set = 1;
314	}
315
316	/* XXX: Callers do not check for failure. */
317	if (!efi_check_space(dest + stage_offset + len)) {
318		errno = ENOMEM;
319		return (-1);
320	}
321	bcopy(src, (void *)(dest + stage_offset), len);
322	return (len);
323}
324
325ssize_t
326efi_copyout(const vm_offset_t src, void *dest, const size_t len)
327{
328
329	/* XXX: Callers do not check for failure. */
330	if (src + stage_offset + len > staging_end) {
331		errno = ENOMEM;
332		return (-1);
333	}
334	bcopy((void *)(src + stage_offset), dest, len);
335	return (len);
336}
337
338
339ssize_t
340efi_readin(readin_handle_t fd, vm_offset_t dest, const size_t len)
341{
342
343	if (!stage_offset_set) {
344		stage_offset = (vm_offset_t)staging - dest;
345		stage_offset_set = 1;
346	}
347
348	if (!efi_check_space(dest + stage_offset + len)) {
349		errno = ENOMEM;
350		return (-1);
351	}
352	return (VECTX_READ(fd, (void *)(dest + stage_offset), len));
353}
354
355void
356efi_copy_finish(void)
357{
358	uint64_t	*src, *dst, *last;
359
360	src = (uint64_t *)(uintptr_t)staging;
361	dst = (uint64_t *)(uintptr_t)(staging - stage_offset);
362	last = (uint64_t *)(uintptr_t)staging_end;
363
364	while (src < last)
365		*dst++ = *src++;
366}
367