2 * (C) Copyright 2002 Linus Torvalds
3 * Portions based on the vdso-randomization code from exec-shield:
4 * Copyright(C) 2005-2006, Red Hat, Inc., Ingo Molnar
6 * This file contains the needed initializations to support sysenter.
9 #include <linux/init.h>
10 #include <linux/smp.h>
11 #include <linux/thread_info.h>
12 #include <linux/sched.h>
13 #include <linux/gfp.h>
14 #include <linux/string.h>
15 #include <linux/elf.h>
17 #include <linux/err.h>
18 #include <linux/module.h>
19 #include <linux/slab.h>
21 #include <asm/cpufeature.h>
23 #include <asm/pgtable.h>
24 #include <asm/unistd.h>
26 #include <asm/tlbflush.h>
28 #include <asm/proto.h>
29 #include <asm/fixmap.h>
33 #ifdef CONFIG_COMPAT_VDSO
34 #define VDSO_DEFAULT 0
36 #define VDSO_DEFAULT 1
40 #define vdso_enabled sysctl_vsyscall32
41 #define arch_setup_additional_pages syscall32_setup_pages
42 extern int sysctl_ldt16;
46 * Should the kernel map a VDSO page into processes and pass its
47 * address down to glibc upon exec()?
49 unsigned int __read_mostly vdso_enabled = VDSO_DEFAULT;
51 static int __init vdso_setup(char *s)
53 vdso_enabled = simple_strtoul(s, NULL, 0);
56 pr_warn("vdso32 values other than 0 and 1 are no longer allowed; vdso disabled\n");
62 * For consistency, the argument vdso32=[012] affects the 32-bit vDSO
63 * behavior on both 64-bit and 32-bit kernels.
64 * On 32-bit kernels, vdso=[012] means the same thing.
66 __setup("vdso32=", vdso_setup);
69 __setup_param("vdso=", vdso32_setup, vdso_setup, 0);
71 EXPORT_SYMBOL_GPL(vdso_enabled);
74 static struct page **vdso32_pages;
75 static unsigned vdso32_size;
79 #define vdso32_sysenter() (boot_cpu_has(X86_FEATURE_SYSENTER32))
80 #define vdso32_syscall() (boot_cpu_has(X86_FEATURE_SYSCALL32))
82 /* May not be __init: called during resume */
83 void syscall32_cpu_init(void)
85 /* Load these always in case some future AMD CPU supports
86 SYSENTER from compat mode too. */
87 wrmsrl_safe(MSR_IA32_SYSENTER_CS, (u64)__KERNEL_CS);
88 wrmsrl_safe(MSR_IA32_SYSENTER_ESP, 0ULL);
89 wrmsrl_safe(MSR_IA32_SYSENTER_EIP, (u64)ia32_sysenter_target);
91 wrmsrl(MSR_CSTAR, ia32_cstar_target);
94 #else /* CONFIG_X86_32 */
96 #define vdso32_sysenter() (boot_cpu_has(X86_FEATURE_SEP))
97 #define vdso32_syscall() (0)
99 void enable_sep_cpu(void)
102 struct tss_struct *tss = &per_cpu(init_tss, cpu);
104 if (!boot_cpu_has(X86_FEATURE_SEP)) {
109 tss->x86_tss.ss1 = __KERNEL_CS;
110 tss->x86_tss.sp1 = sizeof(struct tss_struct) + (unsigned long) tss;
111 wrmsr(MSR_IA32_SYSENTER_CS, __KERNEL_CS, 0);
112 wrmsr(MSR_IA32_SYSENTER_ESP, tss->x86_tss.sp1, 0);
113 wrmsr(MSR_IA32_SYSENTER_EIP, (unsigned long) ia32_sysenter_target, 0);
117 #endif /* CONFIG_X86_64 */
119 int __init sysenter_setup(void)
121 char *vdso32_start, *vdso32_end;
125 if (vdso32_syscall()) {
126 vdso32_start = vdso32_syscall_start;
127 vdso32_end = vdso32_syscall_end;
128 vdso32_pages = vdso32_syscall_pages;
131 if (vdso32_sysenter()) {
132 vdso32_start = vdso32_sysenter_start;
133 vdso32_end = vdso32_sysenter_end;
134 vdso32_pages = vdso32_sysenter_pages;
136 vdso32_start = vdso32_int80_start;
137 vdso32_end = vdso32_int80_end;
138 vdso32_pages = vdso32_int80_pages;
141 npages = ((vdso32_end - vdso32_start) + PAGE_SIZE - 1) / PAGE_SIZE;
142 vdso32_size = npages << PAGE_SHIFT;
143 for (i = 0; i < npages; i++)
144 vdso32_pages[i] = virt_to_page(vdso32_start + i*PAGE_SIZE);
146 patch_vdso32(vdso32_start, vdso32_size);
151 /* Setup a VMA at program startup for the vsyscall page */
152 int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
154 struct mm_struct *mm = current->mm;
157 struct vm_area_struct *vma;
158 static struct page *no_pages[] = {NULL};
160 #ifdef CONFIG_X86_X32_ABI
161 if (test_thread_flag(TIF_X32))
162 return x32_setup_additional_pages(bprm, uses_interp);
165 if (vdso_enabled != 1) /* Other values all mean "disabled" */
168 down_write(&mm->mmap_sem);
170 addr = get_unmapped_area(NULL, 0, vdso32_size + VDSO_OFFSET(VDSO_PREV_PAGES), 0, 0);
171 if (IS_ERR_VALUE(addr)) {
176 addr += VDSO_OFFSET(VDSO_PREV_PAGES);
178 current->mm->context.vdso = (void *)addr;
181 * MAYWRITE to allow gdb to COW and set breakpoints
183 ret = install_special_mapping(mm,
187 VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
193 vma = _install_special_mapping(mm,
194 addr - VDSO_OFFSET(VDSO_PREV_PAGES),
195 VDSO_OFFSET(VDSO_PREV_PAGES),
204 ret = remap_pfn_range(vma,
205 addr - VDSO_OFFSET(VDSO_VVAR_PAGE),
206 __pa_symbol(&__vvar_page) >> PAGE_SHIFT,
213 #ifdef CONFIG_HPET_TIMER
215 ret = io_remap_pfn_range(vma,
216 addr - VDSO_OFFSET(VDSO_HPET_PAGE),
217 hpet_address >> PAGE_SHIFT,
219 pgprot_noncached(PAGE_READONLY));
226 current_thread_info()->sysenter_return =
227 VDSO32_SYMBOL(addr, SYSENTER_RETURN);
231 current->mm->context.vdso = NULL;
233 up_write(&mm->mmap_sem);
240 subsys_initcall(sysenter_setup);
243 /* Register vsyscall32 into the ABI table */
244 #include <linux/sysctl.h>
246 static struct ctl_table abi_table2[] = {
248 .procname = "vsyscall32",
249 .data = &sysctl_vsyscall32,
250 .maxlen = sizeof(int),
252 .proc_handler = proc_dointvec
256 .data = &sysctl_ldt16,
257 .maxlen = sizeof(int),
259 .proc_handler = proc_dointvec
264 static struct ctl_table abi_root_table2[] = {
273 static __init int ia32_binfmt_init(void)
275 register_sysctl_table(abi_root_table2);
278 __initcall(ia32_binfmt_init);
281 #else /* CONFIG_X86_32 */
283 const char *arch_vma_name(struct vm_area_struct *vma)
285 if (vma->vm_mm && vma->vm_start == (long)vma->vm_mm->context.vdso)
290 struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
295 int in_gate_area(struct mm_struct *mm, unsigned long addr)
300 int in_gate_area_no_mm(unsigned long addr)
305 #endif /* CONFIG_X86_64 */