xref: /openbmc/qemu/target/i386/hvf/x86_mmu.c (revision 1be82d89)
1 /*
2  * Copyright (C) 2016 Veertu Inc,
3  * Copyright (C) 2017 Google Inc,
4  *
5  * This program is free software; you can redistribute it and/or
6  * modify it under the terms of the GNU Lesser General Public
7  * License as published by the Free Software Foundation; either
8  * version 2 of the License, or (at your option) any later version.
9  *
10  * This program is distributed in the hope that it will be useful,
11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13  * Lesser General Public License for more details.
14  *
15  * You should have received a copy of the GNU Lesser General Public
16  * License along with this program; if not, see <http://www.gnu.org/licenses/>.
17  */
18 
19 #include "qemu/osdep.h"
20 #include "panic.h"
21 #include "qemu-common.h"
22 #include "cpu.h"
23 #include "x86.h"
24 #include "x86_mmu.h"
25 #include "vmcs.h"
26 #include "vmx.h"
27 #include "exec/address-spaces.h"
28 
29 #define pte_present(pte) (pte & PT_PRESENT)
30 #define pte_write_access(pte) (pte & PT_WRITE)
31 #define pte_user_access(pte) (pte & PT_USER)
32 #define pte_exec_access(pte) (!(pte & PT_NX))
33 
34 #define pte_large_page(pte) (pte & PT_PS)
35 #define pte_global_access(pte) (pte & PT_GLOBAL)
36 
37 #define PAE_CR3_MASK                (~0x1fllu)
38 #define LEGACY_CR3_MASK             (0xffffffff)
39 
40 #define LEGACY_PTE_PAGE_MASK        (0xffffffffllu << 12)
41 #define PAE_PTE_PAGE_MASK           ((-1llu << 12) & ((1llu << 52) - 1))
42 #define PAE_PTE_LARGE_PAGE_MASK     ((-1llu << (21)) & ((1llu << 52) - 1))
43 
44 struct gpt_translation {
45     target_ulong  gva;
46     uint64_t gpa;
47     int    err_code;
48     uint64_t pte[5];
49     bool write_access;
50     bool user_access;
51     bool exec_access;
52 };
53 
54 static int gpt_top_level(struct CPUState *cpu, bool pae)
55 {
56     if (!pae) {
57         return 2;
58     }
59     if (x86_is_long_mode(cpu)) {
60         return 4;
61     }
62 
63     return 3;
64 }
65 
66 static inline int gpt_entry(target_ulong addr, int level, bool pae)
67 {
68     int level_shift = pae ? 9 : 10;
69     return (addr >> (level_shift * (level - 1) + 12)) & ((1 << level_shift) - 1);
70 }
71 
72 static inline int pte_size(bool pae)
73 {
74     return pae ? 8 : 4;
75 }
76 
77 
78 static bool get_pt_entry(struct CPUState *cpu, struct gpt_translation *pt,
79                          int level, bool pae)
80 {
81     int index;
82     uint64_t pte = 0;
83     uint64_t page_mask = pae ? PAE_PTE_PAGE_MASK : LEGACY_PTE_PAGE_MASK;
84     uint64_t gpa = pt->pte[level] & page_mask;
85 
86     if (level == 3 && !x86_is_long_mode(cpu)) {
87         gpa = pt->pte[level];
88     }
89 
90     index = gpt_entry(pt->gva, level, pae);
91     address_space_rw(&address_space_memory, gpa + index * pte_size(pae),
92                      MEMTXATTRS_UNSPECIFIED, (uint8_t *)&pte, pte_size(pae), 0);
93 
94     pt->pte[level - 1] = pte;
95 
96     return true;
97 }
98 
99 /* test page table entry */
100 static bool test_pt_entry(struct CPUState *cpu, struct gpt_translation *pt,
101                           int level, bool *is_large, bool pae)
102 {
103     uint64_t pte = pt->pte[level];
104 
105     if (pt->write_access) {
106         pt->err_code |= MMU_PAGE_WT;
107     }
108     if (pt->user_access) {
109         pt->err_code |= MMU_PAGE_US;
110     }
111     if (pt->exec_access) {
112         pt->err_code |= MMU_PAGE_NX;
113     }
114 
115     if (!pte_present(pte)) {
116         return false;
117     }
118 
119     if (pae && !x86_is_long_mode(cpu) && 2 == level) {
120         goto exit;
121     }
122 
123     if (1 == level && pte_large_page(pte)) {
124         pt->err_code |= MMU_PAGE_PT;
125         *is_large = true;
126     }
127     if (!level) {
128         pt->err_code |= MMU_PAGE_PT;
129     }
130 
131     uint32_t cr0 = rvmcs(cpu->hvf_fd, VMCS_GUEST_CR0);
132     /* check protection */
133     if (cr0 & CR0_WP) {
134         if (pt->write_access && !pte_write_access(pte)) {
135             return false;
136         }
137     }
138 
139     if (pt->user_access && !pte_user_access(pte)) {
140         return false;
141     }
142 
143     if (pae && pt->exec_access && !pte_exec_access(pte)) {
144         return false;
145     }
146 
147 exit:
148     /* TODO: check reserved bits */
149     return true;
150 }
151 
152 static inline uint64_t pse_pte_to_page(uint64_t pte)
153 {
154     return ((pte & 0x1fe000) << 19) | (pte & 0xffc00000);
155 }
156 
157 static inline uint64_t large_page_gpa(struct gpt_translation *pt, bool pae)
158 {
159     VM_PANIC_ON(!pte_large_page(pt->pte[1]))
160     /* 2Mb large page  */
161     if (pae) {
162         return (pt->pte[1] & PAE_PTE_LARGE_PAGE_MASK) | (pt->gva & 0x1fffff);
163     }
164 
165     /* 4Mb large page */
166     return pse_pte_to_page(pt->pte[1]) | (pt->gva & 0x3fffff);
167 }
168 
169 
170 
171 static bool walk_gpt(struct CPUState *cpu, target_ulong addr, int err_code,
172                      struct gpt_translation *pt, bool pae)
173 {
174     int top_level, level;
175     bool is_large = false;
176     target_ulong cr3 = rvmcs(cpu->hvf_fd, VMCS_GUEST_CR3);
177     uint64_t page_mask = pae ? PAE_PTE_PAGE_MASK : LEGACY_PTE_PAGE_MASK;
178 
179     memset(pt, 0, sizeof(*pt));
180     top_level = gpt_top_level(cpu, pae);
181 
182     pt->pte[top_level] = pae ? (cr3 & PAE_CR3_MASK) : (cr3 & LEGACY_CR3_MASK);
183     pt->gva = addr;
184     pt->user_access = (err_code & MMU_PAGE_US);
185     pt->write_access = (err_code & MMU_PAGE_WT);
186     pt->exec_access = (err_code & MMU_PAGE_NX);
187 
188     for (level = top_level; level > 0; level--) {
189         get_pt_entry(cpu, pt, level, pae);
190 
191         if (!test_pt_entry(cpu, pt, level - 1, &is_large, pae)) {
192             return false;
193         }
194 
195         if (is_large) {
196             break;
197         }
198     }
199 
200     if (!is_large) {
201         pt->gpa = (pt->pte[0] & page_mask) | (pt->gva & 0xfff);
202     } else {
203         pt->gpa = large_page_gpa(pt, pae);
204     }
205 
206     return true;
207 }
208 
209 
210 bool mmu_gva_to_gpa(struct CPUState *cpu, target_ulong gva, uint64_t *gpa)
211 {
212     bool res;
213     struct gpt_translation pt;
214     int err_code = 0;
215 
216     if (!x86_is_paging_mode(cpu)) {
217         *gpa = gva;
218         return true;
219     }
220 
221     res = walk_gpt(cpu, gva, err_code, &pt, x86_is_pae_enabled(cpu));
222     if (res) {
223         *gpa = pt.gpa;
224         return true;
225     }
226 
227     return false;
228 }
229 
230 void vmx_write_mem(struct CPUState *cpu, target_ulong gva, void *data, int bytes)
231 {
232     uint64_t gpa;
233 
234     while (bytes > 0) {
235         /* copy page */
236         int copy = MIN(bytes, 0x1000 - (gva & 0xfff));
237 
238         if (!mmu_gva_to_gpa(cpu, gva, &gpa)) {
239             VM_PANIC_EX("%s: mmu_gva_to_gpa %llx failed\n", __func__, gva);
240         } else {
241             address_space_rw(&address_space_memory, gpa, MEMTXATTRS_UNSPECIFIED,
242                              data, copy, 1);
243         }
244 
245         bytes -= copy;
246         gva += copy;
247         data += copy;
248     }
249 }
250 
251 void vmx_read_mem(struct CPUState *cpu, void *data, target_ulong gva, int bytes)
252 {
253     uint64_t gpa;
254 
255     while (bytes > 0) {
256         /* copy page */
257         int copy = MIN(bytes, 0x1000 - (gva & 0xfff));
258 
259         if (!mmu_gva_to_gpa(cpu, gva, &gpa)) {
260             VM_PANIC_EX("%s: mmu_gva_to_gpa %llx failed\n", __func__, gva);
261         }
262         address_space_rw(&address_space_memory, gpa, MEMTXATTRS_UNSPECIFIED,
263                          data, copy, 0);
264 
265         bytes -= copy;
266         gva += copy;
267         data += copy;
268     }
269 }
270