1 /*
2 * Copyright 2007 Andi Kleen, SUSE Labs.
3 * Subject to the GPL, v.2
4 *
5 * This contains most of the x86 vDSO kernel-side code.
6 */
7 #include <linux/mm.h>
8 #include <linux/err.h>
9 #include <linux/sched.h>
10 #include <linux/slab.h>
11 #include <linux/init.h>
12 #include <linux/random.h>
13 #include <linux/elf.h>
14 #include <linux/cpu.h>
15 #include <asm/vgtod.h>
16 #include <asm/proto.h>
17 #include <asm/vdso.h>
18 #include <asm/vvar.h>
19 #include <asm/page.h>
20 #include <asm/hpet.h>
21 #include <asm/desc.h>
22
23 #if defined(CONFIG_X86_64)
24 unsigned int __read_mostly vdso64_enabled = 1;
25 #endif
26
init_vdso_image(const struct vdso_image * image)27 void __init init_vdso_image(const struct vdso_image *image)
28 {
29 int i;
30 int npages = (image->size) / PAGE_SIZE;
31
32 BUG_ON(image->size % PAGE_SIZE != 0);
33 for (i = 0; i < npages; i++)
34 image->text_mapping.pages[i] =
35 virt_to_page(image->data + i*PAGE_SIZE);
36
37 apply_alternatives((struct alt_instr *)(image->data + image->alt),
38 (struct alt_instr *)(image->data + image->alt +
39 image->alt_len));
40 }
41
42 struct linux_binprm;
43
44 /*
45 * Put the vdso above the (randomized) stack with another randomized
46 * offset. This way there is no hole in the middle of address space.
47 * To save memory make sure it is still in the same PTE as the stack
48 * top. This doesn't give that many random bits.
49 *
50 * Note that this algorithm is imperfect: the distribution of the vdso
51 * start address within a PMD is biased toward the end.
52 *
53 * Only used for the 64-bit and x32 vdsos.
54 */
vdso_addr(unsigned long start,unsigned len)55 static unsigned long vdso_addr(unsigned long start, unsigned len)
56 {
57 #ifdef CONFIG_X86_32
58 return 0;
59 #else
60 unsigned long addr, end;
61 unsigned offset;
62
63 /*
64 * Round up the start address. It can start out unaligned as a result
65 * of stack start randomization.
66 */
67 start = PAGE_ALIGN(start);
68
69 /* Round the lowest possible end address up to a PMD boundary. */
70 end = (start + len + PMD_SIZE - 1) & PMD_MASK;
71 if (end >= TASK_SIZE_MAX)
72 end = TASK_SIZE_MAX;
73 end -= len;
74
75 if (end > start) {
76 offset = get_random_int() % (((end - start) >> PAGE_SHIFT) + 1);
77 addr = start + (offset << PAGE_SHIFT);
78 } else {
79 addr = start;
80 }
81
82 /*
83 * Forcibly align the final address in case we have a hardware
84 * issue that requires alignment for performance reasons.
85 */
86 addr = align_vdso_addr(addr);
87
88 return addr;
89 #endif
90 }
91
map_vdso(const struct vdso_image * image,bool calculate_addr)92 static int map_vdso(const struct vdso_image *image, bool calculate_addr)
93 {
94 struct mm_struct *mm = current->mm;
95 struct vm_area_struct *vma;
96 unsigned long addr, text_start;
97 int ret = 0;
98 static struct page *no_pages[] = {NULL};
99 static struct vm_special_mapping vvar_mapping = {
100 .name = "[vvar]",
101 .pages = no_pages,
102 };
103
104 if (calculate_addr) {
105 addr = vdso_addr(current->mm->start_stack,
106 image->size - image->sym_vvar_start);
107 } else {
108 addr = 0;
109 }
110
111 down_write(&mm->mmap_sem);
112
113 addr = get_unmapped_area(NULL, addr,
114 image->size - image->sym_vvar_start, 0, 0);
115 if (IS_ERR_VALUE(addr)) {
116 ret = addr;
117 goto up_fail;
118 }
119
120 text_start = addr - image->sym_vvar_start;
121 current->mm->context.vdso = (void __user *)text_start;
122
123 /*
124 * MAYWRITE to allow gdb to COW and set breakpoints
125 */
126 vma = _install_special_mapping(mm,
127 text_start,
128 image->size,
129 VM_READ|VM_EXEC|
130 VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
131 &image->text_mapping);
132
133 if (IS_ERR(vma)) {
134 ret = PTR_ERR(vma);
135 goto up_fail;
136 }
137
138 vma = _install_special_mapping(mm,
139 addr,
140 -image->sym_vvar_start,
141 VM_READ|VM_MAYREAD,
142 &vvar_mapping);
143
144 if (IS_ERR(vma)) {
145 ret = PTR_ERR(vma);
146 goto up_fail;
147 }
148
149 if (image->sym_vvar_page)
150 ret = remap_pfn_range(vma,
151 text_start + image->sym_vvar_page,
152 __pa_symbol(&__vvar_page) >> PAGE_SHIFT,
153 PAGE_SIZE,
154 PAGE_READONLY);
155
156 if (ret)
157 goto up_fail;
158
159 #ifdef CONFIG_HPET_TIMER
160 if (hpet_address && image->sym_hpet_page) {
161 ret = io_remap_pfn_range(vma,
162 text_start + image->sym_hpet_page,
163 hpet_address >> PAGE_SHIFT,
164 PAGE_SIZE,
165 pgprot_noncached(PAGE_READONLY));
166
167 if (ret)
168 goto up_fail;
169 }
170 #endif
171
172 up_fail:
173 if (ret)
174 current->mm->context.vdso = NULL;
175
176 up_write(&mm->mmap_sem);
177 return ret;
178 }
179
180 #if defined(CONFIG_X86_32) || defined(CONFIG_IA32_EMULATION)
load_vdso32(void)181 static int load_vdso32(void)
182 {
183 if (vdso32_enabled != 1) /* Other values all mean "disabled" */
184 return 0;
185
186 return map_vdso(&vdso_image_32, false);
187 }
188 #endif
189
190 #ifdef CONFIG_X86_64
arch_setup_additional_pages(struct linux_binprm * bprm,int uses_interp)191 int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
192 {
193 if (!vdso64_enabled)
194 return 0;
195
196 return map_vdso(&vdso_image_64, true);
197 }
198
199 #ifdef CONFIG_COMPAT
compat_arch_setup_additional_pages(struct linux_binprm * bprm,int uses_interp)200 int compat_arch_setup_additional_pages(struct linux_binprm *bprm,
201 int uses_interp)
202 {
203 #ifdef CONFIG_X86_X32_ABI
204 if (test_thread_flag(TIF_X32)) {
205 if (!vdso64_enabled)
206 return 0;
207
208 return map_vdso(&vdso_image_x32, true);
209 }
210 #endif
211 #ifdef CONFIG_IA32_EMULATION
212 return load_vdso32();
213 #else
214 return 0;
215 #endif
216 }
217 #endif
218 #else
arch_setup_additional_pages(struct linux_binprm * bprm,int uses_interp)219 int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
220 {
221 return load_vdso32();
222 }
223 #endif
224
225 #ifdef CONFIG_X86_64
vdso_setup(char * s)226 static __init int vdso_setup(char *s)
227 {
228 vdso64_enabled = simple_strtoul(s, NULL, 0);
229 return 0;
230 }
231 __setup("vdso=", vdso_setup);
232 #endif
233
234 #ifdef CONFIG_X86_64
vgetcpu_cpu_init(void * arg)235 static void vgetcpu_cpu_init(void *arg)
236 {
237 int cpu = smp_processor_id();
238 struct desc_struct d = { };
239 unsigned long node = 0;
240 #ifdef CONFIG_NUMA
241 node = cpu_to_node(cpu);
242 #endif
243 if (cpu_has(&cpu_data(cpu), X86_FEATURE_RDTSCP))
244 write_rdtscp_aux((node << 12) | cpu);
245
246 /*
247 * Store cpu number in limit so that it can be loaded
248 * quickly in user space in vgetcpu. (12 bits for the CPU
249 * and 8 bits for the node)
250 */
251 d.limit0 = cpu | ((node & 0xf) << 12);
252 d.limit = node >> 4;
253 d.type = 5; /* RO data, expand down, accessed */
254 d.dpl = 3; /* Visible to user code */
255 d.s = 1; /* Not a system segment */
256 d.p = 1; /* Present */
257 d.d = 1; /* 32-bit */
258
259 write_gdt_entry(get_cpu_gdt_table(cpu), GDT_ENTRY_PER_CPU, &d, DESCTYPE_S);
260 }
261
262 static int
vgetcpu_cpu_notifier(struct notifier_block * n,unsigned long action,void * arg)263 vgetcpu_cpu_notifier(struct notifier_block *n, unsigned long action, void *arg)
264 {
265 long cpu = (long)arg;
266
267 if (action == CPU_ONLINE || action == CPU_ONLINE_FROZEN)
268 smp_call_function_single(cpu, vgetcpu_cpu_init, NULL, 1);
269
270 return NOTIFY_DONE;
271 }
272
init_vdso(void)273 static int __init init_vdso(void)
274 {
275 init_vdso_image(&vdso_image_64);
276
277 #ifdef CONFIG_X86_X32_ABI
278 init_vdso_image(&vdso_image_x32);
279 #endif
280
281 cpu_notifier_register_begin();
282
283 on_each_cpu(vgetcpu_cpu_init, NULL, 1);
284 /* notifier priority > KVM */
285 __hotcpu_notifier(vgetcpu_cpu_notifier, 30);
286
287 cpu_notifier_register_done();
288
289 return 0;
290 }
291 subsys_initcall(init_vdso);
292 #endif /* CONFIG_X86_64 */
293