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