Disintegrate asm/system.h for PowerPC
[linux-block.git] / arch / s390 / kernel / vdso.c
CommitLineData
b020632e
MS
1/*
2 * vdso setup for s390
3 *
4 * Copyright IBM Corp. 2008
5 * Author(s): Martin Schwidefsky (schwidefsky@de.ibm.com)
6 *
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.
10 */
11
12#include <linux/module.h>
13#include <linux/errno.h>
14#include <linux/sched.h>
15#include <linux/kernel.h>
16#include <linux/mm.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>
7757591a 25#include <linux/compat.h>
cbb870c8 26#include <asm/asm-offsets.h>
b020632e
MS
27#include <asm/pgtable.h>
28#include <asm/system.h>
29#include <asm/processor.h>
30#include <asm/mmu.h>
31#include <asm/mmu_context.h>
32#include <asm/sections.h>
33#include <asm/vdso.h>
34
b020632e
MS
35#if defined(CONFIG_32BIT) || defined(CONFIG_COMPAT)
36extern char vdso32_start, vdso32_end;
37static void *vdso32_kbase = &vdso32_start;
38static unsigned int vdso32_pages;
39static struct page **vdso32_pagelist;
40#endif
41
42#ifdef CONFIG_64BIT
43extern char vdso64_start, vdso64_end;
44static void *vdso64_kbase = &vdso64_start;
45static unsigned int vdso64_pages;
46static struct page **vdso64_pagelist;
47#endif /* CONFIG_64BIT */
48
49/*
50 * Should the kernel map a VDSO page into processes and pass its
51 * address down to glibc upon exec()?
52 */
53unsigned int __read_mostly vdso_enabled = 1;
54
55static int __init vdso_setup(char *s)
56{
7aa79f94
MS
57 unsigned long val;
58 int rc;
59
60 rc = 0;
61 if (strncmp(s, "on", 3) == 0)
62 vdso_enabled = 1;
63 else if (strncmp(s, "off", 4) == 0)
64 vdso_enabled = 0;
65 else {
66 rc = strict_strtoul(s, 0, &val);
67 vdso_enabled = rc ? 0 : !!val;
68 }
69 return !rc;
b020632e
MS
70}
71__setup("vdso=", vdso_setup);
72
73/*
74 * The vdso data page
75 */
76static union {
77 struct vdso_data data;
78 u8 page[PAGE_SIZE];
abe1ee3a 79} vdso_data_store __page_aligned_data;
b020632e
MS
80struct vdso_data *vdso_data = &vdso_data_store.data;
81
c742b31c
MS
82/*
83 * Setup vdso data page.
84 */
85static void vdso_init_data(struct vdso_data *vd)
86{
14375bc4 87 vd->ectg_available = user_mode != HOME_SPACE_MODE && test_facility(31);
c742b31c
MS
88}
89
90#ifdef CONFIG_64BIT
c742b31c
MS
91/*
92 * Allocate/free per cpu vdso data.
93 */
c742b31c 94#define SEGMENT_ORDER 2
c742b31c 95
8b646bd7 96int vdso_alloc_per_cpu(struct _lowcore *lowcore)
c742b31c
MS
97{
98 unsigned long segment_table, page_table, page_frame;
99 u32 *psal, *aste;
100 int i;
101
102 lowcore->vdso_per_cpu_data = __LC_PASTE;
103
b11b5334 104 if (user_mode == HOME_SPACE_MODE || !vdso_enabled)
c742b31c
MS
105 return 0;
106
107 segment_table = __get_free_pages(GFP_KERNEL, SEGMENT_ORDER);
108 page_table = get_zeroed_page(GFP_KERNEL | GFP_DMA);
109 page_frame = get_zeroed_page(GFP_KERNEL);
110 if (!segment_table || !page_table || !page_frame)
111 goto out;
112
113 clear_table((unsigned long *) segment_table, _SEGMENT_ENTRY_EMPTY,
114 PAGE_SIZE << SEGMENT_ORDER);
115 clear_table((unsigned long *) page_table, _PAGE_TYPE_EMPTY,
116 256*sizeof(unsigned long));
117
118 *(unsigned long *) segment_table = _SEGMENT_ENTRY + page_table;
119 *(unsigned long *) page_table = _PAGE_RO + page_frame;
120
121 psal = (u32 *) (page_table + 256*sizeof(unsigned long));
122 aste = psal + 32;
123
124 for (i = 4; i < 32; i += 4)
125 psal[i] = 0x80000000;
126
127 lowcore->paste[4] = (u32)(addr_t) psal;
128 psal[0] = 0x20000000;
129 psal[2] = (u32)(addr_t) aste;
130 *(unsigned long *) (aste + 2) = segment_table +
131 _ASCE_TABLE_LENGTH + _ASCE_USER_BITS + _ASCE_TYPE_SEGMENT;
132 aste[4] = (u32)(addr_t) psal;
133 lowcore->vdso_per_cpu_data = page_frame;
134
c742b31c
MS
135 return 0;
136
137out:
138 free_page(page_frame);
139 free_page(page_table);
140 free_pages(segment_table, SEGMENT_ORDER);
141 return -ENOMEM;
142}
143
8b646bd7 144void vdso_free_per_cpu(struct _lowcore *lowcore)
c742b31c
MS
145{
146 unsigned long segment_table, page_table, page_frame;
147 u32 *psal, *aste;
148
b11b5334 149 if (user_mode == HOME_SPACE_MODE || !vdso_enabled)
c742b31c
MS
150 return;
151
152 psal = (u32 *)(addr_t) lowcore->paste[4];
153 aste = (u32 *)(addr_t) psal[2];
154 segment_table = *(unsigned long *)(aste + 2) & PAGE_MASK;
155 page_table = *(unsigned long *) segment_table;
156 page_frame = *(unsigned long *) page_table;
157
158 free_page(page_frame);
159 free_page(page_table);
160 free_pages(segment_table, SEGMENT_ORDER);
161}
c742b31c 162
8b646bd7 163static void vdso_init_cr5(void)
c742b31c
MS
164{
165 unsigned long cr5;
166
8b646bd7
MS
167 if (user_mode == HOME_SPACE_MODE || !vdso_enabled)
168 return;
c742b31c
MS
169 cr5 = offsetof(struct _lowcore, paste);
170 __ctl_load(cr5, 5, 5);
171}
c742b31c
MS
172#endif /* CONFIG_64BIT */
173
b020632e
MS
174/*
175 * This is called from binfmt_elf, we create the special vma for the
176 * vDSO and insert it into the mm struct tree
177 */
178int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
179{
180 struct mm_struct *mm = current->mm;
181 struct page **vdso_pagelist;
182 unsigned long vdso_pages;
183 unsigned long vdso_base;
184 int rc;
185
186 if (!vdso_enabled)
187 return 0;
188 /*
189 * Only map the vdso for dynamically linked elf binaries.
190 */
191 if (!uses_interp)
192 return 0;
193
b020632e
MS
194#ifdef CONFIG_64BIT
195 vdso_pagelist = vdso64_pagelist;
196 vdso_pages = vdso64_pages;
197#ifdef CONFIG_COMPAT
7757591a 198 if (is_compat_task()) {
b020632e
MS
199 vdso_pagelist = vdso32_pagelist;
200 vdso_pages = vdso32_pages;
201 }
202#endif
203#else
204 vdso_pagelist = vdso32_pagelist;
205 vdso_pages = vdso32_pages;
206#endif
207
208 /*
209 * vDSO has a problem and was disabled, just don't "enable" it for
210 * the process
211 */
212 if (vdso_pages == 0)
213 return 0;
214
215 current->mm->context.vdso_base = 0;
216
217 /*
218 * pick a base address for the vDSO in process space. We try to put
219 * it at vdso_base which is the "natural" base for it, but we might
220 * fail and end up putting it elsewhere.
221 */
222 down_write(&mm->mmap_sem);
e7828bbd 223 vdso_base = get_unmapped_area(NULL, 0, vdso_pages << PAGE_SHIFT, 0, 0);
b020632e
MS
224 if (IS_ERR_VALUE(vdso_base)) {
225 rc = vdso_base;
226 goto out_up;
227 }
228
930e44fb
HC
229 /*
230 * Put vDSO base into mm struct. We need to do this before calling
231 * install_special_mapping or the perf counter mmap tracking code
232 * will fail to recognise it as a vDSO (since arch_vma_name fails).
233 */
234 current->mm->context.vdso_base = vdso_base;
235
b020632e
MS
236 /*
237 * our vma flags don't have VM_WRITE so by default, the process
238 * isn't allowed to write those pages.
239 * gdb can break that with ptrace interface, and thus trigger COW
240 * on those pages but it's then your responsibility to never do that
241 * on the "data" page of the vDSO or you'll stop getting kernel
242 * updates and your nice userland gettimeofday will be totally dead.
243 * It's fine to use that for setting breakpoints in the vDSO code
244 * pages though
245 *
246 * Make sure the vDSO gets into every core dump.
247 * Dumping its contents makes post-mortem fully interpretable later
248 * without matching up the same kernel and hardware config to see
249 * what PC values meant.
250 */
251 rc = install_special_mapping(mm, vdso_base, vdso_pages << PAGE_SHIFT,
252 VM_READ|VM_EXEC|
253 VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC|
254 VM_ALWAYSDUMP,
255 vdso_pagelist);
256 if (rc)
930e44fb 257 current->mm->context.vdso_base = 0;
b020632e
MS
258out_up:
259 up_write(&mm->mmap_sem);
260 return rc;
261}
262
263const char *arch_vma_name(struct vm_area_struct *vma)
264{
265 if (vma->vm_mm && vma->vm_start == vma->vm_mm->context.vdso_base)
266 return "[vdso]";
267 return NULL;
268}
269
270static int __init vdso_init(void)
271{
272 int i;
273
c742b31c
MS
274 if (!vdso_enabled)
275 return 0;
276 vdso_init_data(vdso_data);
b020632e
MS
277#if defined(CONFIG_32BIT) || defined(CONFIG_COMPAT)
278 /* Calculate the size of the 32 bit vDSO */
279 vdso32_pages = ((&vdso32_end - &vdso32_start
280 + PAGE_SIZE - 1) >> PAGE_SHIFT) + 1;
281
282 /* Make sure pages are in the correct state */
283 vdso32_pagelist = kzalloc(sizeof(struct page *) * (vdso32_pages + 1),
284 GFP_KERNEL);
285 BUG_ON(vdso32_pagelist == NULL);
286 for (i = 0; i < vdso32_pages - 1; i++) {
287 struct page *pg = virt_to_page(vdso32_kbase + i*PAGE_SIZE);
288 ClearPageReserved(pg);
289 get_page(pg);
290 vdso32_pagelist[i] = pg;
291 }
292 vdso32_pagelist[vdso32_pages - 1] = virt_to_page(vdso_data);
293 vdso32_pagelist[vdso32_pages] = NULL;
294#endif
295
296#ifdef CONFIG_64BIT
297 /* Calculate the size of the 64 bit vDSO */
298 vdso64_pages = ((&vdso64_end - &vdso64_start
299 + PAGE_SIZE - 1) >> PAGE_SHIFT) + 1;
300
301 /* Make sure pages are in the correct state */
302 vdso64_pagelist = kzalloc(sizeof(struct page *) * (vdso64_pages + 1),
303 GFP_KERNEL);
304 BUG_ON(vdso64_pagelist == NULL);
305 for (i = 0; i < vdso64_pages - 1; i++) {
306 struct page *pg = virt_to_page(vdso64_kbase + i*PAGE_SIZE);
307 ClearPageReserved(pg);
308 get_page(pg);
309 vdso64_pagelist[i] = pg;
310 }
311 vdso64_pagelist[vdso64_pages - 1] = virt_to_page(vdso_data);
312 vdso64_pagelist[vdso64_pages] = NULL;
8b646bd7 313 if (vdso_alloc_per_cpu(&S390_lowcore))
81ffa041 314 BUG();
c742b31c 315 vdso_init_cr5();
b020632e
MS
316#endif /* CONFIG_64BIT */
317
318 get_page(virt_to_page(vdso_data));
319
320 smp_wmb();
321
322 return 0;
323}
8b646bd7 324early_initcall(vdso_init);
b020632e 325
cae5d390 326int in_gate_area_no_mm(unsigned long addr)
b020632e
MS
327{
328 return 0;
329}
330
83b964bb 331int in_gate_area(struct mm_struct *mm, unsigned long addr)
b020632e
MS
332{
333 return 0;
334}
335
31db58b3 336struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
b020632e
MS
337{
338 return NULL;
339}