1 // SPDX-License-Identifier: GPL-2.0
5 * Copyright IBM Corp. 2008
6 * Author(s): Martin Schwidefsky (schwidefsky@de.ibm.com)
9 #include <linux/init.h>
10 #include <linux/errno.h>
11 #include <linux/sched.h>
12 #include <linux/kernel.h>
14 #include <linux/smp.h>
15 #include <linux/stddef.h>
16 #include <linux/unistd.h>
17 #include <linux/slab.h>
18 #include <linux/user.h>
19 #include <linux/elf.h>
20 #include <linux/security.h>
21 #include <linux/memblock.h>
22 #include <linux/compat.h>
23 #include <asm/asm-offsets.h>
24 #include <asm/processor.h>
26 #include <asm/mmu_context.h>
27 #include <asm/sections.h>
29 #include <asm/facility.h>
31 extern char vdso64_start, vdso64_end;
32 static void *vdso64_kbase = &vdso64_start;
33 static unsigned int vdso64_pages;
34 static struct page **vdso64_pagelist;
37 * Should the kernel map a VDSO page into processes and pass its
38 * address down to glibc upon exec()?
40 unsigned int __read_mostly vdso_enabled = 1;
42 static vm_fault_t vdso_fault(const struct vm_special_mapping *sm,
43 struct vm_area_struct *vma, struct vm_fault *vmf)
45 struct page **vdso_pagelist;
46 unsigned long vdso_pages;
48 vdso_pagelist = vdso64_pagelist;
49 vdso_pages = vdso64_pages;
51 if (vmf->pgoff >= vdso_pages)
52 return VM_FAULT_SIGBUS;
54 vmf->page = vdso_pagelist[vmf->pgoff];
59 static int vdso_mremap(const struct vm_special_mapping *sm,
60 struct vm_area_struct *vma)
62 unsigned long vdso_pages;
64 vdso_pages = vdso64_pages;
66 if ((vdso_pages << PAGE_SHIFT) != vma->vm_end - vma->vm_start)
69 if (WARN_ON_ONCE(current->mm != vma->vm_mm))
72 current->mm->context.vdso_base = vma->vm_start;
76 static const struct vm_special_mapping vdso_mapping = {
79 .mremap = vdso_mremap,
82 static int __init vdso_setup(char *str)
86 if (!kstrtobool(str, &enabled))
87 vdso_enabled = enabled;
90 __setup("vdso=", vdso_setup);
96 struct vdso_data data;
98 } vdso_data_store __page_aligned_data;
99 struct vdso_data *vdso_data = &vdso_data_store.data;
102 * Setup vdso data page.
104 static void __init vdso_init_data(struct vdso_data *vd)
106 vd->ectg_available = test_facility(31);
110 * Allocate/free per cpu vdso data.
112 #define SEGMENT_ORDER 2
115 * The initial vdso_data structure for the boot CPU. Eventually
116 * it is replaced with a properly allocated structure in vdso_init.
117 * This is necessary because a valid S390_lowcore.vdso_per_cpu_data
118 * pointer is required to be able to return from an interrupt or
119 * program check. See the exit paths in entry.S.
121 struct vdso_data boot_vdso_data __initdata;
123 void __init vdso_alloc_boot_cpu(struct lowcore *lowcore)
125 lowcore->vdso_per_cpu_data = (unsigned long) &boot_vdso_data;
128 int vdso_alloc_per_cpu(struct lowcore *lowcore)
130 unsigned long segment_table, page_table, page_frame;
131 struct vdso_per_cpu_data *vd;
133 segment_table = __get_free_pages(GFP_KERNEL, SEGMENT_ORDER);
134 page_table = get_zeroed_page(GFP_KERNEL);
135 page_frame = get_zeroed_page(GFP_KERNEL);
136 if (!segment_table || !page_table || !page_frame)
138 arch_set_page_dat(virt_to_page(segment_table), SEGMENT_ORDER);
139 arch_set_page_dat(virt_to_page(page_table), 0);
141 /* Initialize per-cpu vdso data page */
142 vd = (struct vdso_per_cpu_data *) page_frame;
143 vd->cpu_nr = lowcore->cpu_nr;
144 vd->node_id = cpu_to_node(vd->cpu_nr);
146 /* Set up page table for the vdso address space */
147 memset64((u64 *)segment_table, _SEGMENT_ENTRY_EMPTY, _CRST_ENTRIES);
148 memset64((u64 *)page_table, _PAGE_INVALID, PTRS_PER_PTE);
150 *(unsigned long *) segment_table = _SEGMENT_ENTRY + page_table;
151 *(unsigned long *) page_table = _PAGE_PROTECT + page_frame;
153 lowcore->vdso_asce = segment_table +
154 _ASCE_TABLE_LENGTH + _ASCE_USER_BITS + _ASCE_TYPE_SEGMENT;
155 lowcore->vdso_per_cpu_data = page_frame;
160 free_page(page_frame);
161 free_page(page_table);
162 free_pages(segment_table, SEGMENT_ORDER);
166 void vdso_free_per_cpu(struct lowcore *lowcore)
168 unsigned long segment_table, page_table, page_frame;
170 segment_table = lowcore->vdso_asce & PAGE_MASK;
171 page_table = *(unsigned long *) segment_table;
172 page_frame = *(unsigned long *) page_table;
174 free_page(page_frame);
175 free_page(page_table);
176 free_pages(segment_table, SEGMENT_ORDER);
180 * This is called from binfmt_elf, we create the special vma for the
181 * vDSO and insert it into the mm struct tree
183 int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
185 struct mm_struct *mm = current->mm;
186 struct vm_area_struct *vma;
187 unsigned long vdso_pages;
188 unsigned long vdso_base;
194 if (is_compat_task())
197 vdso_pages = vdso64_pages;
199 * vDSO has a problem and was disabled, just don't "enable" it for
206 * pick a base address for the vDSO in process space. We try to put
207 * it at vdso_base which is the "natural" base for it, but we might
208 * fail and end up putting it elsewhere.
210 if (mmap_write_lock_killable(mm))
212 vdso_base = get_unmapped_area(NULL, 0, vdso_pages << PAGE_SHIFT, 0, 0);
213 if (IS_ERR_VALUE(vdso_base)) {
219 * our vma flags don't have VM_WRITE so by default, the process
220 * isn't allowed to write those pages.
221 * gdb can break that with ptrace interface, and thus trigger COW
222 * on those pages but it's then your responsibility to never do that
223 * on the "data" page of the vDSO or you'll stop getting kernel
224 * updates and your nice userland gettimeofday will be totally dead.
225 * It's fine to use that for setting breakpoints in the vDSO code
228 vma = _install_special_mapping(mm, vdso_base, vdso_pages << PAGE_SHIFT,
230 VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
237 current->mm->context.vdso_base = vdso_base;
241 mmap_write_unlock(mm);
245 static int __init vdso_init(void)
249 vdso_init_data(vdso_data);
251 /* Calculate the size of the 64 bit vDSO */
252 vdso64_pages = ((&vdso64_end - &vdso64_start
253 + PAGE_SIZE - 1) >> PAGE_SHIFT) + 1;
255 /* Make sure pages are in the correct state */
256 vdso64_pagelist = kcalloc(vdso64_pages + 1, sizeof(struct page *),
258 BUG_ON(vdso64_pagelist == NULL);
259 for (i = 0; i < vdso64_pages - 1; i++) {
260 struct page *pg = virt_to_page(vdso64_kbase + i*PAGE_SIZE);
262 vdso64_pagelist[i] = pg;
264 vdso64_pagelist[vdso64_pages - 1] = virt_to_page(vdso_data);
265 vdso64_pagelist[vdso64_pages] = NULL;
266 if (vdso_alloc_per_cpu(&S390_lowcore))
269 get_page(virt_to_page(vdso_data));
273 early_initcall(vdso_init);