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