1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3 * Copyright 2007 Andi Kleen, SUSE Labs.
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/sched/task_stack.h>
11 #include <linux/slab.h>
12 #include <linux/init.h>
13 #include <linux/random.h>
14 #include <linux/elf.h>
15 #include <linux/cpu.h>
16 #include <linux/ptrace.h>
17 #include <linux/vdso_datastore.h>
18
19 #include <asm/pvclock.h>
20 #include <asm/vgtod.h>
21 #include <asm/proto.h>
22 #include <asm/vdso.h>
23 #include <asm/tlb.h>
24 #include <asm/page.h>
25 #include <asm/desc.h>
26 #include <asm/cpufeature.h>
27 #include <asm/vdso/vsyscall.h>
28 #include <clocksource/hyperv_timer.h>
29
30 static_assert(VDSO_NR_PAGES + VDSO_NR_VCLOCK_PAGES == __VDSO_PAGES);
31
32 unsigned int vclocks_used __read_mostly;
33
34 #if defined(CONFIG_X86_64)
35 unsigned int __read_mostly vdso64_enabled = 1;
36 #endif
37
init_vdso_image(const struct vdso_image * image)38 int __init init_vdso_image(const struct vdso_image *image)
39 {
40 BUILD_BUG_ON(VDSO_CLOCKMODE_MAX >= 32);
41 BUG_ON(image->size % PAGE_SIZE != 0);
42
43 apply_alternatives((struct alt_instr *)(image->data + image->alt),
44 (struct alt_instr *)(image->data + image->alt +
45 image->alt_len));
46
47 return 0;
48 }
49
50 struct linux_binprm;
51
vdso_fault(const struct vm_special_mapping * sm,struct vm_area_struct * vma,struct vm_fault * vmf)52 static vm_fault_t vdso_fault(const struct vm_special_mapping *sm,
53 struct vm_area_struct *vma, struct vm_fault *vmf)
54 {
55 const struct vdso_image *image = vma->vm_mm->context.vdso_image;
56
57 if (!image || (vmf->pgoff << PAGE_SHIFT) >= image->size)
58 return VM_FAULT_SIGBUS;
59
60 vmf->page = virt_to_page(image->data + (vmf->pgoff << PAGE_SHIFT));
61 get_page(vmf->page);
62 return 0;
63 }
64
vdso_fix_landing(const struct vdso_image * image,struct vm_area_struct * new_vma)65 static void vdso_fix_landing(const struct vdso_image *image,
66 struct vm_area_struct *new_vma)
67 {
68 struct pt_regs *regs = current_pt_regs();
69 unsigned long ipoffset = regs->ip -
70 (unsigned long)current->mm->context.vdso;
71
72 if (ipoffset < image->size)
73 regs->ip = new_vma->vm_start + ipoffset;
74 }
75
vdso_mremap(const struct vm_special_mapping * sm,struct vm_area_struct * new_vma)76 static int vdso_mremap(const struct vm_special_mapping *sm,
77 struct vm_area_struct *new_vma)
78 {
79 const struct vdso_image *image = current->mm->context.vdso_image;
80
81 vdso_fix_landing(image, new_vma);
82 current->mm->context.vdso = (void __user *)new_vma->vm_start;
83
84 return 0;
85 }
86
vvar_vclock_fault(const struct vm_special_mapping * sm,struct vm_area_struct * vma,struct vm_fault * vmf)87 static vm_fault_t vvar_vclock_fault(const struct vm_special_mapping *sm,
88 struct vm_area_struct *vma, struct vm_fault *vmf)
89 {
90 switch (vmf->pgoff) {
91 #ifdef CONFIG_PARAVIRT_CLOCK
92 case VDSO_PAGE_PVCLOCK_OFFSET:
93 {
94 struct pvclock_vsyscall_time_info *pvti =
95 pvclock_get_pvti_cpu0_va();
96
97 if (pvti && vclock_was_used(VDSO_CLOCKMODE_PVCLOCK))
98 return vmf_insert_pfn_prot(vma, vmf->address,
99 __pa(pvti) >> PAGE_SHIFT,
100 pgprot_decrypted(vma->vm_page_prot));
101 break;
102 }
103 #endif /* CONFIG_PARAVIRT_CLOCK */
104 #ifdef CONFIG_HYPERV_TIMER
105 case VDSO_PAGE_HVCLOCK_OFFSET:
106 {
107 unsigned long pfn = hv_get_tsc_pfn();
108 if (pfn && vclock_was_used(VDSO_CLOCKMODE_HVCLOCK))
109 return vmf_insert_pfn(vma, vmf->address, pfn);
110 break;
111 }
112 #endif /* CONFIG_HYPERV_TIMER */
113 }
114
115 return VM_FAULT_SIGBUS;
116 }
117
118 static const struct vm_special_mapping vdso_mapping = {
119 .name = "[vdso]",
120 .fault = vdso_fault,
121 .mremap = vdso_mremap,
122 };
123 static const struct vm_special_mapping vvar_vclock_mapping = {
124 .name = "[vvar_vclock]",
125 .fault = vvar_vclock_fault,
126 };
127
128 /*
129 * Add vdso and vvar mappings to current process.
130 * @image - blob to map
131 * @addr - request a specific address (zero to map at free addr)
132 */
map_vdso(const struct vdso_image * image,unsigned long addr)133 static int map_vdso(const struct vdso_image *image, unsigned long addr)
134 {
135 struct mm_struct *mm = current->mm;
136 struct vm_area_struct *vma;
137 unsigned long text_start;
138 int ret = 0;
139
140 if (mmap_write_lock_killable(mm))
141 return -EINTR;
142
143 addr = get_unmapped_area(NULL, addr,
144 image->size + __VDSO_PAGES * PAGE_SIZE, 0, 0);
145 if (IS_ERR_VALUE(addr)) {
146 ret = addr;
147 goto up_fail;
148 }
149
150 text_start = addr + __VDSO_PAGES * PAGE_SIZE;
151
152 /*
153 * MAYWRITE to allow gdb to COW and set breakpoints
154 */
155 vma = _install_special_mapping(mm,
156 text_start,
157 image->size,
158 VM_READ|VM_EXEC|
159 VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC|
160 VM_SEALED_SYSMAP,
161 &vdso_mapping);
162
163 if (IS_ERR(vma)) {
164 ret = PTR_ERR(vma);
165 goto up_fail;
166 }
167
168 vma = vdso_install_vvar_mapping(mm, addr);
169 if (IS_ERR(vma)) {
170 ret = PTR_ERR(vma);
171 do_munmap(mm, text_start, image->size, NULL);
172 goto up_fail;
173 }
174
175 vma = _install_special_mapping(mm,
176 VDSO_VCLOCK_PAGES_START(addr),
177 VDSO_NR_VCLOCK_PAGES * PAGE_SIZE,
178 VM_READ|VM_MAYREAD|VM_IO|VM_DONTDUMP|
179 VM_PFNMAP|VM_SEALED_SYSMAP,
180 &vvar_vclock_mapping);
181
182 if (IS_ERR(vma)) {
183 ret = PTR_ERR(vma);
184 do_munmap(mm, text_start, image->size, NULL);
185 do_munmap(mm, addr, image->size, NULL);
186 goto up_fail;
187 }
188
189 current->mm->context.vdso = (void __user *)text_start;
190 current->mm->context.vdso_image = image;
191
192 up_fail:
193 mmap_write_unlock(mm);
194 return ret;
195 }
196
map_vdso_once(const struct vdso_image * image,unsigned long addr)197 int map_vdso_once(const struct vdso_image *image, unsigned long addr)
198 {
199 struct mm_struct *mm = current->mm;
200 struct vm_area_struct *vma;
201 VMA_ITERATOR(vmi, mm, 0);
202
203 mmap_write_lock(mm);
204 /*
205 * Check if we have already mapped vdso blob - fail to prevent
206 * abusing from userspace install_special_mapping, which may
207 * not do accounting and rlimit right.
208 * We could search vma near context.vdso, but it's a slowpath,
209 * so let's explicitly check all VMAs to be completely sure.
210 */
211 for_each_vma(vmi, vma) {
212 if (vma_is_special_mapping(vma, &vdso_mapping) ||
213 vma_is_special_mapping(vma, &vdso_vvar_mapping) ||
214 vma_is_special_mapping(vma, &vvar_vclock_mapping)) {
215 mmap_write_unlock(mm);
216 return -EEXIST;
217 }
218 }
219 mmap_write_unlock(mm);
220
221 return map_vdso(image, addr);
222 }
223
load_vdso32(void)224 static int load_vdso32(void)
225 {
226 if (vdso32_enabled != 1) /* Other values all mean "disabled" */
227 return 0;
228
229 return map_vdso(&vdso32_image, 0);
230 }
231
arch_setup_additional_pages(struct linux_binprm * bprm,int uses_interp)232 int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
233 {
234 if (IS_ENABLED(CONFIG_X86_64)) {
235 if (!vdso64_enabled)
236 return 0;
237
238 return map_vdso(&vdso64_image, 0);
239 }
240
241 return load_vdso32();
242 }
243
244 #ifdef CONFIG_COMPAT
compat_arch_setup_additional_pages(struct linux_binprm * bprm,int uses_interp,bool x32)245 int compat_arch_setup_additional_pages(struct linux_binprm *bprm,
246 int uses_interp, bool x32)
247 {
248 if (IS_ENABLED(CONFIG_X86_X32_ABI) && x32) {
249 if (!vdso64_enabled)
250 return 0;
251 return map_vdso(&vdsox32_image, 0);
252 }
253
254 if (IS_ENABLED(CONFIG_IA32_EMULATION))
255 return load_vdso32();
256
257 return 0;
258 }
259 #endif
260
arch_syscall_is_vdso_sigreturn(struct pt_regs * regs)261 bool arch_syscall_is_vdso_sigreturn(struct pt_regs *regs)
262 {
263 const struct vdso_image *image = current->mm->context.vdso_image;
264 unsigned long vdso = (unsigned long) current->mm->context.vdso;
265
266 if (in_ia32_syscall() && image == &vdso32_image) {
267 if (regs->ip == vdso + image->sym_vdso32_sigreturn_landing_pad ||
268 regs->ip == vdso + image->sym_vdso32_rt_sigreturn_landing_pad)
269 return true;
270 }
271 return false;
272 }
273
274 #ifdef CONFIG_X86_64
vdso_setup(char * s)275 static __init int vdso_setup(char *s)
276 {
277 vdso64_enabled = simple_strtoul(s, NULL, 0);
278 return 1;
279 }
280 __setup("vdso=", vdso_setup);
281 #endif /* CONFIG_X86_64 */
282