arch/s390/kernel/vdso.c

   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>
  25 #include <linux/compat.h>
  26 #include <asm/pgtable.h>
  27 #include <asm/system.h>
  28 #include <asm/processor.h>
  29 #include <asm/mmu.h>
  30 #include <asm/mmu_context.h>
  31 #include <asm/sections.h>
  32 #include <asm/vdso.h>
  33
  34 #if defined(CONFIG_32BIT) || defined(CONFIG_COMPAT)
  35 extern char vdso32_start, vdso32_end;
  36 static void *vdso32_kbase = &vdso32_start;
  37 static unsigned int vdso32_pages;
  38 static struct page **vdso32_pagelist;
  39 #endif
  40
  41 #ifdef CONFIG_64BIT
  42 extern char vdso64_start, vdso64_end;
  43 static void *vdso64_kbase = &vdso64_start;
  44 static unsigned int vdso64_pages;
  45 static struct page **vdso64_pagelist;
  46 #endif /* CONFIG_64BIT */
  47
  48 /*
  49  * Should the kernel map a VDSO page into processes and pass its
  50  * address down to glibc upon exec()?
  51  */
  52 unsigned int __read_mostly vdso_enabled = 1;
  53
  54 static int __init vdso_setup(char *s)
  55 {
  56         unsigned long val;
  57         int rc;
  58
  59         rc = 0;
  60         if (strncmp(s, "on", 3) == 0)
  61                 vdso_enabled = 1;
  62         else if (strncmp(s, "off", 4) == 0)
  63                 vdso_enabled = 0;
  64         else {
  65                 rc = strict_strtoul(s, 0, &val);
  66                 vdso_enabled = rc ? 0 : !!val;
  67         }
  68         return !rc;
  69 }
  70 __setup("vdso=", vdso_setup);
  71
  72 /*
  73  * The vdso data page
  74  */
  75 static union {
  76         struct vdso_data        data;
  77         u8                      page[PAGE_SIZE];
  78 } vdso_data_store __page_aligned_data;
  79 struct vdso_data *vdso_data = &vdso_data_store.data;
  80
  81 /*
  82  * Setup vdso data page.
  83  */
  84 static void vdso_init_data(struct vdso_data *vd)
  85 {
  86         unsigned int facility_list;
  87
  88         facility_list = stfl();
  89         vd->ectg_available =
  90                 user_mode != HOME_SPACE_MODE && (facility_list & 1);
  91 }
  92
  93 #ifdef CONFIG_64BIT
  94 /*
  95  * Setup per cpu vdso data page.
  96  */
  97 static void vdso_init_per_cpu_data(int cpu, struct vdso_per_cpu_data *vpcd)
  98 {
  99 }
 100
 101 /*
 102  * Allocate/free per cpu vdso data.
 103  */
 104 #ifdef CONFIG_64BIT
 105 #define SEGMENT_ORDER   2
 106 #else
 107 #define SEGMENT_ORDER   1
 108 #endif
 109
 110 int vdso_alloc_per_cpu(int cpu, struct _lowcore *lowcore)
 111 {
 112         unsigned long segment_table, page_table, page_frame;
 113         u32 *psal, *aste;
 114         int i;
 115
 116         lowcore->vdso_per_cpu_data = __LC_PASTE;
 117
 118         if (user_mode == HOME_SPACE_MODE || !vdso_enabled)
 119                 return 0;
 120
 121         segment_table = __get_free_pages(GFP_KERNEL, SEGMENT_ORDER);
 122         page_table = get_zeroed_page(GFP_KERNEL | GFP_DMA);
 123         page_frame = get_zeroed_page(GFP_KERNEL);
 124         if (!segment_table || !page_table || !page_frame)
 125                 goto out;
 126
 127         clear_table((unsigned long *) segment_table, _SEGMENT_ENTRY_EMPTY,
 128                     PAGE_SIZE << SEGMENT_ORDER);
 129         clear_table((unsigned long *) page_table, _PAGE_TYPE_EMPTY,
 130                     256*sizeof(unsigned long));
 131
 132         *(unsigned long *) segment_table = _SEGMENT_ENTRY + page_table;
 133         *(unsigned long *) page_table = _PAGE_RO + page_frame;
 134
 135         psal = (u32 *) (page_table + 256*sizeof(unsigned long));
 136         aste = psal + 32;
 137
 138         for (i = 4; i < 32; i += 4)
 139                 psal[i] = 0x80000000;
 140
 141         lowcore->paste[4] = (u32)(addr_t) psal;
 142         psal[0] = 0x20000000;
 143         psal[2] = (u32)(addr_t) aste;
 144         *(unsigned long *) (aste + 2) = segment_table +
 145                 _ASCE_TABLE_LENGTH + _ASCE_USER_BITS + _ASCE_TYPE_SEGMENT;
 146         aste[4] = (u32)(addr_t) psal;
 147         lowcore->vdso_per_cpu_data = page_frame;
 148
 149         vdso_init_per_cpu_data(cpu, (struct vdso_per_cpu_data *) page_frame);
 150         return 0;
 151
 152 out:
 153         free_page(page_frame);
 154         free_page(page_table);
 155         free_pages(segment_table, SEGMENT_ORDER);
 156         return -ENOMEM;
 157 }
 158
 159 void vdso_free_per_cpu(int cpu, struct _lowcore *lowcore)
 160 {
 161         unsigned long segment_table, page_table, page_frame;
 162         u32 *psal, *aste;
 163
 164         if (user_mode == HOME_SPACE_MODE || !vdso_enabled)
 165                 return;
 166
 167         psal = (u32 *)(addr_t) lowcore->paste[4];
 168         aste = (u32 *)(addr_t) psal[2];
 169         segment_table = *(unsigned long *)(aste + 2) & PAGE_MASK;
 170         page_table = *(unsigned long *) segment_table;
 171         page_frame = *(unsigned long *) page_table;
 172
 173         free_page(page_frame);
 174         free_page(page_table);
 175         free_pages(segment_table, SEGMENT_ORDER);
 176 }
 177
 178 static void __vdso_init_cr5(void *dummy)
 179 {
 180         unsigned long cr5;
 181
 182         cr5 = offsetof(struct _lowcore, paste);
 183         __ctl_load(cr5, 5, 5);
 184 }
 185
 186 static void vdso_init_cr5(void)
 187 {
 188         if (user_mode != HOME_SPACE_MODE && vdso_enabled)
 189                 on_each_cpu(__vdso_init_cr5, NULL, 1);
 190 }
 191 #endif /* CONFIG_64BIT */
 192
 193 /*
 194  * This is called from binfmt_elf, we create the special vma for the
 195  * vDSO and insert it into the mm struct tree
 196  */
 197 int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 198 {
 199         struct mm_struct *mm = current->mm;
 200         struct page **vdso_pagelist;
 201         unsigned long vdso_pages;
 202         unsigned long vdso_base;
 203         int rc;
 204
 205         if (!vdso_enabled)
 206                 return 0;
 207         /*
 208          * Only map the vdso for dynamically linked elf binaries.
 209          */
 210         if (!uses_interp)
 211                 return 0;
 212
 213         vdso_base = mm->mmap_base;
 214 #ifdef CONFIG_64BIT
 215         vdso_pagelist = vdso64_pagelist;
 216         vdso_pages = vdso64_pages;
 217 #ifdef CONFIG_COMPAT
 218         if (is_compat_task()) {
 219                 vdso_pagelist = vdso32_pagelist;
 220                 vdso_pages = vdso32_pages;
 221         }
 222 #endif
 223 #else
 224         vdso_pagelist = vdso32_pagelist;
 225         vdso_pages = vdso32_pages;
 226 #endif
 227
 228         /*
 229          * vDSO has a problem and was disabled, just don't "enable" it for
 230          * the process
 231          */
 232         if (vdso_pages == 0)
 233                 return 0;
 234
 235         current->mm->context.vdso_base = 0;
 236
 237         /*
 238          * pick a base address for the vDSO in process space. We try to put
 239          * it at vdso_base which is the "natural" base for it, but we might
 240          * fail and end up putting it elsewhere.
 241          */
 242         down_write(&mm->mmap_sem);
 243         vdso_base = get_unmapped_area(NULL, vdso_base,
 244                                       vdso_pages << PAGE_SHIFT, 0, 0);
 245         if (IS_ERR_VALUE(vdso_base)) {
 246                 rc = vdso_base;
 247                 goto out_up;
 248         }
 249
 250         /*
 251          * Put vDSO base into mm struct. We need to do this before calling
 252          * install_special_mapping or the perf counter mmap tracking code
 253          * will fail to recognise it as a vDSO (since arch_vma_name fails).
 254          */
 255         current->mm->context.vdso_base = vdso_base;
 256
 257         /*
 258          * our vma flags don't have VM_WRITE so by default, the process
 259          * isn't allowed to write those pages.
 260          * gdb can break that with ptrace interface, and thus trigger COW
 261          * on those pages but it's then your responsibility to never do that
 262          * on the "data" page of the vDSO or you'll stop getting kernel
 263          * updates and your nice userland gettimeofday will be totally dead.
 264          * It's fine to use that for setting breakpoints in the vDSO code
 265          * pages though
 266          *
 267          * Make sure the vDSO gets into every core dump.
 268          * Dumping its contents makes post-mortem fully interpretable later
 269          * without matching up the same kernel and hardware config to see
 270          * what PC values meant.
 271          */
 272         rc = install_special_mapping(mm, vdso_base, vdso_pages << PAGE_SHIFT,
 273                                      VM_READ|VM_EXEC|
 274                                      VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC|
 275                                      VM_ALWAYSDUMP,
 276                                      vdso_pagelist);
 277         if (rc)
 278                 current->mm->context.vdso_base = 0;
 279 out_up:
 280         up_write(&mm->mmap_sem);
 281         return rc;
 282 }
 283
 284 const char *arch_vma_name(struct vm_area_struct *vma)
 285 {
 286         if (vma->vm_mm && vma->vm_start == vma->vm_mm->context.vdso_base)
 287                 return "[vdso]";
 288         return NULL;
 289 }
 290
 291 static int __init vdso_init(void)
 292 {
 293         int i;
 294
 295         if (!vdso_enabled)
 296                 return 0;
 297         vdso_init_data(vdso_data);
 298 #if defined(CONFIG_32BIT) || defined(CONFIG_COMPAT)
 299         /* Calculate the size of the 32 bit vDSO */
 300         vdso32_pages = ((&vdso32_end - &vdso32_start
 301                          + PAGE_SIZE - 1) >> PAGE_SHIFT) + 1;
 302
 303         /* Make sure pages are in the correct state */
 304         vdso32_pagelist = kzalloc(sizeof(struct page *) * (vdso32_pages + 1),
 305                                   GFP_KERNEL);
 306         BUG_ON(vdso32_pagelist == NULL);
 307         for (i = 0; i < vdso32_pages - 1; i++) {
 308                 struct page *pg = virt_to_page(vdso32_kbase + i*PAGE_SIZE);
 309                 ClearPageReserved(pg);
 310                 get_page(pg);
 311                 vdso32_pagelist[i] = pg;
 312         }
 313         vdso32_pagelist[vdso32_pages - 1] = virt_to_page(vdso_data);
 314         vdso32_pagelist[vdso32_pages] = NULL;
 315 #endif
 316
 317 #ifdef CONFIG_64BIT
 318         /* Calculate the size of the 64 bit vDSO */
 319         vdso64_pages = ((&vdso64_end - &vdso64_start
 320                          + PAGE_SIZE - 1) >> PAGE_SHIFT) + 1;
 321
 322         /* Make sure pages are in the correct state */
 323         vdso64_pagelist = kzalloc(sizeof(struct page *) * (vdso64_pages + 1),
 324                                   GFP_KERNEL);
 325         BUG_ON(vdso64_pagelist == NULL);
 326         for (i = 0; i < vdso64_pages - 1; i++) {
 327                 struct page *pg = virt_to_page(vdso64_kbase + i*PAGE_SIZE);
 328                 ClearPageReserved(pg);
 329                 get_page(pg);
 330                 vdso64_pagelist[i] = pg;
 331         }
 332         vdso64_pagelist[vdso64_pages - 1] = virt_to_page(vdso_data);
 333         vdso64_pagelist[vdso64_pages] = NULL;
 334 #ifndef CONFIG_SMP
 335         if (vdso_alloc_per_cpu(0, &S390_lowcore))
 336                 BUG();
 337 #endif
 338         vdso_init_cr5();
 339 #endif /* CONFIG_64BIT */
 340
 341         get_page(virt_to_page(vdso_data));
 342
 343         smp_wmb();
 344
 345         return 0;
 346 }
 347 arch_initcall(vdso_init);
 348
 349 int in_gate_area_no_task(unsigned long addr)
 350 {
 351         return 0;
 352 }
 353
 354 int in_gate_area(struct task_struct *task, unsigned long addr)
 355 {
 356         return 0;
 357 }
 358
 359 struct vm_area_struct *get_gate_vma(struct task_struct *tsk)
 360 {
 361         return NULL;
 362 }