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 = kstrtoul(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 = test_facility(31);
  88 }
  89
  90 #ifdef CONFIG_64BIT
  91 /*
  92  * Allocate/free per cpu vdso data.
  93  */
  94 #define SEGMENT_ORDER   2
  95
  96 int vdso_alloc_per_cpu(struct _lowcore *lowcore)
  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
 104         if (!vdso_enabled)
 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_INVALID,
 116                     256*sizeof(unsigned long));
 117
 118         *(unsigned long *) segment_table = _SEGMENT_ENTRY + page_table;
 119         *(unsigned long *) page_table = _PAGE_PROTECT + 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] = 0x02000000;
 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
 135         return 0;
 136
 137 out:
 138         free_page(page_frame);
 139         free_page(page_table);
 140         free_pages(segment_table, SEGMENT_ORDER);
 141         return -ENOMEM;
 142 }
 143
 144 void vdso_free_per_cpu(struct _lowcore *lowcore)
 145 {
 146         unsigned long segment_table, page_table, page_frame;
 147         u32 *psal, *aste;
 148
 149         if (!vdso_enabled)
 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 }
 162
 163 static void vdso_init_cr5(void)
 164 {
 165         unsigned long cr5;
 166
 167         if (!vdso_enabled)
 168                 return;
 169         cr5 = offsetof(struct _lowcore, paste);
 170         __ctl_load(cr5, 5, 5);
 171 }
 172 #endif /* CONFIG_64BIT */
 173
 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  */
 178 int 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
 194 #ifdef CONFIG_64BIT
 195         vdso_pagelist = vdso64_pagelist;
 196         vdso_pages = vdso64_pages;
 197 #ifdef CONFIG_COMPAT
 198         if (is_compat_task()) {
 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);
 223         vdso_base = get_unmapped_area(NULL, 0, vdso_pages << PAGE_SHIFT, 0, 0);
 224         if (IS_ERR_VALUE(vdso_base)) {
 225                 rc = vdso_base;
 226                 goto out_up;
 227         }
 228
 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
 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         rc = install_special_mapping(mm, vdso_base, vdso_pages << PAGE_SHIFT,
 247                                      VM_READ|VM_EXEC|
 248                                      VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
 249                                      vdso_pagelist);
 250         if (rc)
 251                 current->mm->context.vdso_base = 0;
 252 out_up:
 253         up_write(&mm->mmap_sem);
 254         return rc;
 255 }
 256
 257 const char *arch_vma_name(struct vm_area_struct *vma)
 258 {
 259         if (vma->vm_mm && vma->vm_start == vma->vm_mm->context.vdso_base)
 260                 return "[vdso]";
 261         return NULL;
 262 }
 263
 264 static int __init vdso_init(void)
 265 {
 266         int i;
 267
 268         if (!vdso_enabled)
 269                 return 0;
 270         vdso_init_data(vdso_data);
 271 #if defined(CONFIG_32BIT) || defined(CONFIG_COMPAT)
 272         /* Calculate the size of the 32 bit vDSO */
 273         vdso32_pages = ((&vdso32_end - &vdso32_start
 274                          + PAGE_SIZE - 1) >> PAGE_SHIFT) + 1;
 275
 276         /* Make sure pages are in the correct state */
 277         vdso32_pagelist = kzalloc(sizeof(struct page *) * (vdso32_pages + 1),
 278                                   GFP_KERNEL);
 279         BUG_ON(vdso32_pagelist == NULL);
 280         for (i = 0; i < vdso32_pages - 1; i++) {
 281                 struct page *pg = virt_to_page(vdso32_kbase + i*PAGE_SIZE);
 282                 ClearPageReserved(pg);
 283                 get_page(pg);
 284                 vdso32_pagelist[i] = pg;
 285         }
 286         vdso32_pagelist[vdso32_pages - 1] = virt_to_page(vdso_data);
 287         vdso32_pagelist[vdso32_pages] = NULL;
 288 #endif
 289
 290 #ifdef CONFIG_64BIT
 291         /* Calculate the size of the 64 bit vDSO */
 292         vdso64_pages = ((&vdso64_end - &vdso64_start
 293                          + PAGE_SIZE - 1) >> PAGE_SHIFT) + 1;
 294
 295         /* Make sure pages are in the correct state */
 296         vdso64_pagelist = kzalloc(sizeof(struct page *) * (vdso64_pages + 1),
 297                                   GFP_KERNEL);
 298         BUG_ON(vdso64_pagelist == NULL);
 299         for (i = 0; i < vdso64_pages - 1; i++) {
 300                 struct page *pg = virt_to_page(vdso64_kbase + i*PAGE_SIZE);
 301                 ClearPageReserved(pg);
 302                 get_page(pg);
 303                 vdso64_pagelist[i] = pg;
 304         }
 305         vdso64_pagelist[vdso64_pages - 1] = virt_to_page(vdso_data);
 306         vdso64_pagelist[vdso64_pages] = NULL;
 307         if (vdso_alloc_per_cpu(&S390_lowcore))
 308                 BUG();
 309         vdso_init_cr5();
 310 #endif /* CONFIG_64BIT */
 311
 312         get_page(virt_to_page(vdso_data));
 313
 314         smp_wmb();
 315
 316         return 0;
 317 }
 318 early_initcall(vdso_init);