4 * Copyright IBM Corp. 2008
5 * Author(s): Martin Schwidefsky (schwidefsky@de.ibm.com)
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License (version 2 only)
9 * as published by the Free Software Foundation.
12 #include <linux/module.h>
13 #include <linux/errno.h>
14 #include <linux/sched.h>
15 #include <linux/kernel.h>
17 #include <linux/smp.h>
18 #include <linux/stddef.h>
19 #include <linux/unistd.h>
20 #include <linux/slab.h>
21 #include <linux/user.h>
22 #include <linux/elf.h>
23 #include <linux/security.h>
24 #include <linux/bootmem.h>
25 #include <linux/compat.h>
26 #include <asm/asm-offsets.h>
27 #include <asm/pgtable.h>
28 #include <asm/system.h>
29 #include <asm/processor.h>
31 #include <asm/mmu_context.h>
32 #include <asm/sections.h>
35 #if defined(CONFIG_32BIT) || defined(CONFIG_COMPAT)
36 extern char vdso32_start, vdso32_end;
37 static void *vdso32_kbase = &vdso32_start;
38 static unsigned int vdso32_pages;
39 static struct page **vdso32_pagelist;
43 extern char vdso64_start, vdso64_end;
44 static void *vdso64_kbase = &vdso64_start;
45 static unsigned int vdso64_pages;
46 static struct page **vdso64_pagelist;
47 #endif /* CONFIG_64BIT */
50 * Should the kernel map a VDSO page into processes and pass its
51 * address down to glibc upon exec()?
53 unsigned int __read_mostly vdso_enabled = 1;
55 static int __init vdso_setup(char *s)
61 if (strncmp(s, "on", 3) == 0)
63 else if (strncmp(s, "off", 4) == 0)
66 rc = strict_strtoul(s, 0, &val);
67 vdso_enabled = rc ? 0 : !!val;
71 __setup("vdso=", vdso_setup);
77 struct vdso_data data;
79 } vdso_data_store __page_aligned_data;
80 struct vdso_data *vdso_data = &vdso_data_store.data;
83 * Setup vdso data page.
85 static void vdso_init_data(struct vdso_data *vd)
87 unsigned int facility_list;
89 facility_list = stfl();
91 user_mode != HOME_SPACE_MODE && (facility_list & 1);
96 * Setup per cpu vdso data page.
98 static void vdso_init_per_cpu_data(int cpu, struct vdso_per_cpu_data *vpcd)
103 * Allocate/free per cpu vdso data.
106 #define SEGMENT_ORDER 2
108 #define SEGMENT_ORDER 1
111 int vdso_alloc_per_cpu(int cpu, struct _lowcore *lowcore)
113 unsigned long segment_table, page_table, page_frame;
117 lowcore->vdso_per_cpu_data = __LC_PASTE;
119 if (user_mode == HOME_SPACE_MODE || !vdso_enabled)
122 segment_table = __get_free_pages(GFP_KERNEL, SEGMENT_ORDER);
123 page_table = get_zeroed_page(GFP_KERNEL | GFP_DMA);
124 page_frame = get_zeroed_page(GFP_KERNEL);
125 if (!segment_table || !page_table || !page_frame)
128 clear_table((unsigned long *) segment_table, _SEGMENT_ENTRY_EMPTY,
129 PAGE_SIZE << SEGMENT_ORDER);
130 clear_table((unsigned long *) page_table, _PAGE_TYPE_EMPTY,
131 256*sizeof(unsigned long));
133 *(unsigned long *) segment_table = _SEGMENT_ENTRY + page_table;
134 *(unsigned long *) page_table = _PAGE_RO + page_frame;
136 psal = (u32 *) (page_table + 256*sizeof(unsigned long));
139 for (i = 4; i < 32; i += 4)
140 psal[i] = 0x80000000;
142 lowcore->paste[4] = (u32)(addr_t) psal;
143 psal[0] = 0x20000000;
144 psal[2] = (u32)(addr_t) aste;
145 *(unsigned long *) (aste + 2) = segment_table +
146 _ASCE_TABLE_LENGTH + _ASCE_USER_BITS + _ASCE_TYPE_SEGMENT;
147 aste[4] = (u32)(addr_t) psal;
148 lowcore->vdso_per_cpu_data = page_frame;
150 vdso_init_per_cpu_data(cpu, (struct vdso_per_cpu_data *) page_frame);
154 free_page(page_frame);
155 free_page(page_table);
156 free_pages(segment_table, SEGMENT_ORDER);
160 void vdso_free_per_cpu(int cpu, struct _lowcore *lowcore)
162 unsigned long segment_table, page_table, page_frame;
165 if (user_mode == HOME_SPACE_MODE || !vdso_enabled)
168 psal = (u32 *)(addr_t) lowcore->paste[4];
169 aste = (u32 *)(addr_t) psal[2];
170 segment_table = *(unsigned long *)(aste + 2) & 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);
179 static void __vdso_init_cr5(void *dummy)
183 cr5 = offsetof(struct _lowcore, paste);
184 __ctl_load(cr5, 5, 5);
187 static void vdso_init_cr5(void)
189 if (user_mode != HOME_SPACE_MODE && vdso_enabled)
190 on_each_cpu(__vdso_init_cr5, NULL, 1);
192 #endif /* CONFIG_64BIT */
195 * This is called from binfmt_elf, we create the special vma for the
196 * vDSO and insert it into the mm struct tree
198 int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
200 struct mm_struct *mm = current->mm;
201 struct page **vdso_pagelist;
202 unsigned long vdso_pages;
203 unsigned long vdso_base;
209 * Only map the vdso for dynamically linked elf binaries.
214 vdso_base = mm->mmap_base;
216 vdso_pagelist = vdso64_pagelist;
217 vdso_pages = vdso64_pages;
219 if (is_compat_task()) {
220 vdso_pagelist = vdso32_pagelist;
221 vdso_pages = vdso32_pages;
225 vdso_pagelist = vdso32_pagelist;
226 vdso_pages = vdso32_pages;
230 * vDSO has a problem and was disabled, just don't "enable" it for
236 current->mm->context.vdso_base = 0;
239 * pick a base address for the vDSO in process space. We try to put
240 * it at vdso_base which is the "natural" base for it, but we might
241 * fail and end up putting it elsewhere.
243 down_write(&mm->mmap_sem);
244 vdso_base = get_unmapped_area(NULL, vdso_base,
245 vdso_pages << PAGE_SHIFT, 0, 0);
246 if (IS_ERR_VALUE(vdso_base)) {
252 * Put vDSO base into mm struct. We need to do this before calling
253 * install_special_mapping or the perf counter mmap tracking code
254 * will fail to recognise it as a vDSO (since arch_vma_name fails).
256 current->mm->context.vdso_base = vdso_base;
259 * our vma flags don't have VM_WRITE so by default, the process
260 * isn't allowed to write those pages.
261 * gdb can break that with ptrace interface, and thus trigger COW
262 * on those pages but it's then your responsibility to never do that
263 * on the "data" page of the vDSO or you'll stop getting kernel
264 * updates and your nice userland gettimeofday will be totally dead.
265 * It's fine to use that for setting breakpoints in the vDSO code
268 * Make sure the vDSO gets into every core dump.
269 * Dumping its contents makes post-mortem fully interpretable later
270 * without matching up the same kernel and hardware config to see
271 * what PC values meant.
273 rc = install_special_mapping(mm, vdso_base, vdso_pages << PAGE_SHIFT,
275 VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC|
279 current->mm->context.vdso_base = 0;
281 up_write(&mm->mmap_sem);
285 const char *arch_vma_name(struct vm_area_struct *vma)
287 if (vma->vm_mm && vma->vm_start == vma->vm_mm->context.vdso_base)
292 static int __init vdso_init(void)
298 vdso_init_data(vdso_data);
299 #if defined(CONFIG_32BIT) || defined(CONFIG_COMPAT)
300 /* Calculate the size of the 32 bit vDSO */
301 vdso32_pages = ((&vdso32_end - &vdso32_start
302 + PAGE_SIZE - 1) >> PAGE_SHIFT) + 1;
304 /* Make sure pages are in the correct state */
305 vdso32_pagelist = kzalloc(sizeof(struct page *) * (vdso32_pages + 1),
307 BUG_ON(vdso32_pagelist == NULL);
308 for (i = 0; i < vdso32_pages - 1; i++) {
309 struct page *pg = virt_to_page(vdso32_kbase + i*PAGE_SIZE);
310 ClearPageReserved(pg);
312 vdso32_pagelist[i] = pg;
314 vdso32_pagelist[vdso32_pages - 1] = virt_to_page(vdso_data);
315 vdso32_pagelist[vdso32_pages] = NULL;
319 /* Calculate the size of the 64 bit vDSO */
320 vdso64_pages = ((&vdso64_end - &vdso64_start
321 + PAGE_SIZE - 1) >> PAGE_SHIFT) + 1;
323 /* Make sure pages are in the correct state */
324 vdso64_pagelist = kzalloc(sizeof(struct page *) * (vdso64_pages + 1),
326 BUG_ON(vdso64_pagelist == NULL);
327 for (i = 0; i < vdso64_pages - 1; i++) {
328 struct page *pg = virt_to_page(vdso64_kbase + i*PAGE_SIZE);
329 ClearPageReserved(pg);
331 vdso64_pagelist[i] = pg;
333 vdso64_pagelist[vdso64_pages - 1] = virt_to_page(vdso_data);
334 vdso64_pagelist[vdso64_pages] = NULL;
336 if (vdso_alloc_per_cpu(0, &S390_lowcore))
340 #endif /* CONFIG_64BIT */
342 get_page(virt_to_page(vdso_data));
348 arch_initcall(vdso_init);
350 int in_gate_area_no_task(unsigned long addr)
355 int in_gate_area(struct task_struct *task, unsigned long addr)
360 struct vm_area_struct *get_gate_vma(struct task_struct *tsk)