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
  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         vdso_enabled = simple_strtoul(s, NULL, 0);
  57         return 1;
  58 }
  59 __setup("vdso=", vdso_setup);
  60
  61 /*
  62  * The vdso data page
  63  */
  64 static union {
  65         struct vdso_data        data;
  66         u8                      page[PAGE_SIZE];
  67 } vdso_data_store __attribute__((__section__(".data.page_aligned")));
  68 struct vdso_data *vdso_data = &vdso_data_store.data;
  69
  70 /*
  71  * Setup vdso data page.
  72  */
  73 static void vdso_init_data(struct vdso_data *vd)
  74 {
  75         unsigned int facility_list;
  76
  77         facility_list = stfl();
  78         vd->ectg_available = switch_amode && (facility_list & 1);
  79 }
  80
  81 #ifdef CONFIG_64BIT
  82 /*
  83  * Setup per cpu vdso data page.
  84  */
  85 static void vdso_init_per_cpu_data(int cpu, struct vdso_per_cpu_data *vpcd)
  86 {
  87 }
  88
  89 /*
  90  * Allocate/free per cpu vdso data.
  91  */
  92 #ifdef CONFIG_64BIT
  93 #define SEGMENT_ORDER   2
  94 #else
  95 #define SEGMENT_ORDER   1
  96 #endif
  97
  98 int vdso_alloc_per_cpu(int cpu, struct _lowcore *lowcore)
  99 {
 100         unsigned long segment_table, page_table, page_frame;
 101         u32 *psal, *aste;
 102         int i;
 103
 104         lowcore->vdso_per_cpu_data = __LC_PASTE;
 105
 106         if (!switch_amode || !vdso_enabled)
 107                 return 0;
 108
 109         segment_table = __get_free_pages(GFP_KERNEL, SEGMENT_ORDER);
 110         page_table = get_zeroed_page(GFP_KERNEL | GFP_DMA);
 111         page_frame = get_zeroed_page(GFP_KERNEL);
 112         if (!segment_table || !page_table || !page_frame)
 113                 goto out;
 114
 115         clear_table((unsigned long *) segment_table, _SEGMENT_ENTRY_EMPTY,
 116                     PAGE_SIZE << SEGMENT_ORDER);
 117         clear_table((unsigned long *) page_table, _PAGE_TYPE_EMPTY,
 118                     256*sizeof(unsigned long));
 119
 120         *(unsigned long *) segment_table = _SEGMENT_ENTRY + page_table;
 121         *(unsigned long *) page_table = _PAGE_RO + page_frame;
 122
 123         psal = (u32 *) (page_table + 256*sizeof(unsigned long));
 124         aste = psal + 32;
 125
 126         for (i = 4; i < 32; i += 4)
 127                 psal[i] = 0x80000000;
 128
 129         lowcore->paste[4] = (u32)(addr_t) psal;
 130         psal[0] = 0x20000000;
 131         psal[2] = (u32)(addr_t) aste;
 132         *(unsigned long *) (aste + 2) = segment_table +
 133                 _ASCE_TABLE_LENGTH + _ASCE_USER_BITS + _ASCE_TYPE_SEGMENT;
 134         aste[4] = (u32)(addr_t) psal;
 135         lowcore->vdso_per_cpu_data = page_frame;
 136
 137         vdso_init_per_cpu_data(cpu, (struct vdso_per_cpu_data *) page_frame);
 138         return 0;
 139
 140 out:
 141         free_page(page_frame);
 142         free_page(page_table);
 143         free_pages(segment_table, SEGMENT_ORDER);
 144         return -ENOMEM;
 145 }
 146
 147 #ifdef CONFIG_HOTPLUG_CPU
 148 void vdso_free_per_cpu(int cpu, struct _lowcore *lowcore)
 149 {
 150         unsigned long segment_table, page_table, page_frame;
 151         u32 *psal, *aste;
 152
 153         if (!switch_amode || !vdso_enabled)
 154                 return;
 155
 156         psal = (u32 *)(addr_t) lowcore->paste[4];
 157         aste = (u32 *)(addr_t) psal[2];
 158         segment_table = *(unsigned long *)(aste + 2) & PAGE_MASK;
 159         page_table = *(unsigned long *) segment_table;
 160         page_frame = *(unsigned long *) page_table;
 161
 162         free_page(page_frame);
 163         free_page(page_table);
 164         free_pages(segment_table, SEGMENT_ORDER);
 165 }
 166 #endif /* CONFIG_HOTPLUG_CPU */
 167
 168 static void __vdso_init_cr5(void *dummy)
 169 {
 170         unsigned long cr5;
 171
 172         cr5 = offsetof(struct _lowcore, paste);
 173         __ctl_load(cr5, 5, 5);
 174 }
 175
 176 static void vdso_init_cr5(void)
 177 {
 178         if (switch_amode && vdso_enabled)
 179                 on_each_cpu(__vdso_init_cr5, NULL, 1);
 180 }
 181 #endif /* CONFIG_64BIT */
 182
 183 /*
 184  * This is called from binfmt_elf, we create the special vma for the
 185  * vDSO and insert it into the mm struct tree
 186  */
 187 int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 188 {
 189         struct mm_struct *mm = current->mm;
 190         struct page **vdso_pagelist;
 191         unsigned long vdso_pages;
 192         unsigned long vdso_base;
 193         int rc;
 194
 195         if (!vdso_enabled)
 196                 return 0;
 197         /*
 198          * Only map the vdso for dynamically linked elf binaries.
 199          */
 200         if (!uses_interp)
 201                 return 0;
 202
 203         vdso_base = mm->mmap_base;
 204 #ifdef CONFIG_64BIT
 205         vdso_pagelist = vdso64_pagelist;
 206         vdso_pages = vdso64_pages;
 207 #ifdef CONFIG_COMPAT
 208         if (test_thread_flag(TIF_31BIT)) {
 209                 vdso_pagelist = vdso32_pagelist;
 210                 vdso_pages = vdso32_pages;
 211         }
 212 #endif
 213 #else
 214         vdso_pagelist = vdso32_pagelist;
 215         vdso_pages = vdso32_pages;
 216 #endif
 217
 218         /*
 219          * vDSO has a problem and was disabled, just don't "enable" it for
 220          * the process
 221          */
 222         if (vdso_pages == 0)
 223                 return 0;
 224
 225         current->mm->context.vdso_base = 0;
 226
 227         /*
 228          * pick a base address for the vDSO in process space. We try to put
 229          * it at vdso_base which is the "natural" base for it, but we might
 230          * fail and end up putting it elsewhere.
 231          */
 232         down_write(&mm->mmap_sem);
 233         vdso_base = get_unmapped_area(NULL, vdso_base,
 234                                       vdso_pages << PAGE_SHIFT, 0, 0);
 235         if (IS_ERR_VALUE(vdso_base)) {
 236                 rc = vdso_base;
 237                 goto out_up;
 238         }
 239
 240         /*
 241          * our vma flags don't have VM_WRITE so by default, the process
 242          * isn't allowed to write those pages.
 243          * gdb can break that with ptrace interface, and thus trigger COW
 244          * on those pages but it's then your responsibility to never do that
 245          * on the "data" page of the vDSO or you'll stop getting kernel
 246          * updates and your nice userland gettimeofday will be totally dead.
 247          * It's fine to use that for setting breakpoints in the vDSO code
 248          * pages though
 249          *
 250          * Make sure the vDSO gets into every core dump.
 251          * Dumping its contents makes post-mortem fully interpretable later
 252          * without matching up the same kernel and hardware config to see
 253          * what PC values meant.
 254          */
 255         rc = install_special_mapping(mm, vdso_base, vdso_pages << PAGE_SHIFT,
 256                                      VM_READ|VM_EXEC|
 257                                      VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC|
 258                                      VM_ALWAYSDUMP,
 259                                      vdso_pagelist);
 260         if (rc)
 261                 goto out_up;
 262
 263         /* Put vDSO base into mm struct */
 264         current->mm->context.vdso_base = vdso_base;
 265
 266         up_write(&mm->mmap_sem);
 267         return 0;
 268
 269 out_up:
 270         up_write(&mm->mmap_sem);
 271         return rc;
 272 }
 273
 274 const char *arch_vma_name(struct vm_area_struct *vma)
 275 {
 276         if (vma->vm_mm && vma->vm_start == vma->vm_mm->context.vdso_base)
 277                 return "[vdso]";
 278         return NULL;
 279 }
 280
 281 static int __init vdso_init(void)
 282 {
 283         int i;
 284
 285         if (!vdso_enabled)
 286                 return 0;
 287         vdso_init_data(vdso_data);
 288 #if defined(CONFIG_32BIT) || defined(CONFIG_COMPAT)
 289         /* Calculate the size of the 32 bit vDSO */
 290         vdso32_pages = ((&vdso32_end - &vdso32_start
 291                          + PAGE_SIZE - 1) >> PAGE_SHIFT) + 1;
 292
 293         /* Make sure pages are in the correct state */
 294         vdso32_pagelist = kzalloc(sizeof(struct page *) * (vdso32_pages + 1),
 295                                   GFP_KERNEL);
 296         BUG_ON(vdso32_pagelist == NULL);
 297         for (i = 0; i < vdso32_pages - 1; i++) {
 298                 struct page *pg = virt_to_page(vdso32_kbase + i*PAGE_SIZE);
 299                 ClearPageReserved(pg);
 300                 get_page(pg);
 301                 vdso32_pagelist[i] = pg;
 302         }
 303         vdso32_pagelist[vdso32_pages - 1] = virt_to_page(vdso_data);
 304         vdso32_pagelist[vdso32_pages] = NULL;
 305 #endif
 306
 307 #ifdef CONFIG_64BIT
 308         /* Calculate the size of the 64 bit vDSO */
 309         vdso64_pages = ((&vdso64_end - &vdso64_start
 310                          + PAGE_SIZE - 1) >> PAGE_SHIFT) + 1;
 311
 312         /* Make sure pages are in the correct state */
 313         vdso64_pagelist = kzalloc(sizeof(struct page *) * (vdso64_pages + 1),
 314                                   GFP_KERNEL);
 315         BUG_ON(vdso64_pagelist == NULL);
 316         for (i = 0; i < vdso64_pages - 1; i++) {
 317                 struct page *pg = virt_to_page(vdso64_kbase + i*PAGE_SIZE);
 318                 ClearPageReserved(pg);
 319                 get_page(pg);
 320                 vdso64_pagelist[i] = pg;
 321         }
 322         vdso64_pagelist[vdso64_pages - 1] = virt_to_page(vdso_data);
 323         vdso64_pagelist[vdso64_pages] = NULL;
 324 #ifndef CONFIG_SMP
 325         if (vdso_alloc_per_cpu(0, &S390_lowcore))
 326                 BUG();
 327 #endif
 328         vdso_init_cr5();
 329 #endif /* CONFIG_64BIT */
 330
 331         get_page(virt_to_page(vdso_data));
 332
 333         smp_wmb();
 334
 335         return 0;
 336 }
 337 arch_initcall(vdso_init);
 338
 339 int in_gate_area_no_task(unsigned long addr)
 340 {
 341         return 0;
 342 }
 343
 344 int in_gate_area(struct task_struct *task, unsigned long addr)
 345 {
 346         return 0;
 347 }
 348
 349 struct vm_area_struct *get_gate_vma(struct task_struct *tsk)
 350 {
 351         return NULL;
 352 }