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 void vdso_free_per_cpu(int cpu, struct _lowcore *lowcore)
 148 {
 149         unsigned long segment_table, page_table, page_frame;
 150         u32 *psal, *aste;
 151
 152         if (!switch_amode || !vdso_enabled)
 153                 return;
 154
 155         psal = (u32 *)(addr_t) lowcore->paste[4];
 156         aste = (u32 *)(addr_t) psal[2];
 157         segment_table = *(unsigned long *)(aste + 2) & PAGE_MASK;
 158         page_table = *(unsigned long *) segment_table;
 159         page_frame = *(unsigned long *) page_table;
 160
 161         free_page(page_frame);
 162         free_page(page_table);
 163         free_pages(segment_table, SEGMENT_ORDER);
 164 }
 165
 166 static void __vdso_init_cr5(void *dummy)
 167 {
 168         unsigned long cr5;
 169
 170         cr5 = offsetof(struct _lowcore, paste);
 171         __ctl_load(cr5, 5, 5);
 172 }
 173
 174 static void vdso_init_cr5(void)
 175 {
 176         if (switch_amode && vdso_enabled)
 177                 on_each_cpu(__vdso_init_cr5, NULL, 1);
 178 }
 179 #endif /* CONFIG_64BIT */
 180
 181 /*
 182  * This is called from binfmt_elf, we create the special vma for the
 183  * vDSO and insert it into the mm struct tree
 184  */
 185 int arch_setup_additional_pages(struct linux_binprm *bprm, int uses_interp)
 186 {
 187         struct mm_struct *mm = current->mm;
 188         struct page **vdso_pagelist;
 189         unsigned long vdso_pages;
 190         unsigned long vdso_base;
 191         int rc;
 192
 193         if (!vdso_enabled)
 194                 return 0;
 195         /*
 196          * Only map the vdso for dynamically linked elf binaries.
 197          */
 198         if (!uses_interp)
 199                 return 0;
 200
 201         vdso_base = mm->mmap_base;
 202 #ifdef CONFIG_64BIT
 203         vdso_pagelist = vdso64_pagelist;
 204         vdso_pages = vdso64_pages;
 205 #ifdef CONFIG_COMPAT
 206         if (test_thread_flag(TIF_31BIT)) {
 207                 vdso_pagelist = vdso32_pagelist;
 208                 vdso_pages = vdso32_pages;
 209         }
 210 #endif
 211 #else
 212         vdso_pagelist = vdso32_pagelist;
 213         vdso_pages = vdso32_pages;
 214 #endif
 215
 216         /*
 217          * vDSO has a problem and was disabled, just don't "enable" it for
 218          * the process
 219          */
 220         if (vdso_pages == 0)
 221                 return 0;
 222
 223         current->mm->context.vdso_base = 0;
 224
 225         /*
 226          * pick a base address for the vDSO in process space. We try to put
 227          * it at vdso_base which is the "natural" base for it, but we might
 228          * fail and end up putting it elsewhere.
 229          */
 230         down_write(&mm->mmap_sem);
 231         vdso_base = get_unmapped_area(NULL, vdso_base,
 232                                       vdso_pages << PAGE_SHIFT, 0, 0);
 233         if (IS_ERR_VALUE(vdso_base)) {
 234                 rc = vdso_base;
 235                 goto out_up;
 236         }
 237
 238         /*
 239          * our vma flags don't have VM_WRITE so by default, the process
 240          * isn't allowed to write those pages.
 241          * gdb can break that with ptrace interface, and thus trigger COW
 242          * on those pages but it's then your responsibility to never do that
 243          * on the "data" page of the vDSO or you'll stop getting kernel
 244          * updates and your nice userland gettimeofday will be totally dead.
 245          * It's fine to use that for setting breakpoints in the vDSO code
 246          * pages though
 247          *
 248          * Make sure the vDSO gets into every core dump.
 249          * Dumping its contents makes post-mortem fully interpretable later
 250          * without matching up the same kernel and hardware config to see
 251          * what PC values meant.
 252          */
 253         rc = install_special_mapping(mm, vdso_base, vdso_pages << PAGE_SHIFT,
 254                                      VM_READ|VM_EXEC|
 255                                      VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC|
 256                                      VM_ALWAYSDUMP,
 257                                      vdso_pagelist);
 258         if (rc)
 259                 goto out_up;
 260
 261         /* Put vDSO base into mm struct */
 262         current->mm->context.vdso_base = vdso_base;
 263
 264         up_write(&mm->mmap_sem);
 265         return 0;
 266
 267 out_up:
 268         up_write(&mm->mmap_sem);
 269         return rc;
 270 }
 271
 272 const char *arch_vma_name(struct vm_area_struct *vma)
 273 {
 274         if (vma->vm_mm && vma->vm_start == vma->vm_mm->context.vdso_base)
 275                 return "[vdso]";
 276         return NULL;
 277 }
 278
 279 static int __init vdso_init(void)
 280 {
 281         int i;
 282
 283         if (!vdso_enabled)
 284                 return 0;
 285         vdso_init_data(vdso_data);
 286 #if defined(CONFIG_32BIT) || defined(CONFIG_COMPAT)
 287         /* Calculate the size of the 32 bit vDSO */
 288         vdso32_pages = ((&vdso32_end - &vdso32_start
 289                          + PAGE_SIZE - 1) >> PAGE_SHIFT) + 1;
 290
 291         /* Make sure pages are in the correct state */
 292         vdso32_pagelist = kzalloc(sizeof(struct page *) * (vdso32_pages + 1),
 293                                   GFP_KERNEL);
 294         BUG_ON(vdso32_pagelist == NULL);
 295         for (i = 0; i < vdso32_pages - 1; i++) {
 296                 struct page *pg = virt_to_page(vdso32_kbase + i*PAGE_SIZE);
 297                 ClearPageReserved(pg);
 298                 get_page(pg);
 299                 vdso32_pagelist[i] = pg;
 300         }
 301         vdso32_pagelist[vdso32_pages - 1] = virt_to_page(vdso_data);
 302         vdso32_pagelist[vdso32_pages] = NULL;
 303 #endif
 304
 305 #ifdef CONFIG_64BIT
 306         /* Calculate the size of the 64 bit vDSO */
 307         vdso64_pages = ((&vdso64_end - &vdso64_start
 308                          + PAGE_SIZE - 1) >> PAGE_SHIFT) + 1;
 309
 310         /* Make sure pages are in the correct state */
 311         vdso64_pagelist = kzalloc(sizeof(struct page *) * (vdso64_pages + 1),
 312                                   GFP_KERNEL);
 313         BUG_ON(vdso64_pagelist == NULL);
 314         for (i = 0; i < vdso64_pages - 1; i++) {
 315                 struct page *pg = virt_to_page(vdso64_kbase + i*PAGE_SIZE);
 316                 ClearPageReserved(pg);
 317                 get_page(pg);
 318                 vdso64_pagelist[i] = pg;
 319         }
 320         vdso64_pagelist[vdso64_pages - 1] = virt_to_page(vdso_data);
 321         vdso64_pagelist[vdso64_pages] = NULL;
 322 #ifndef CONFIG_SMP
 323         if (vdso_alloc_per_cpu(0, &S390_lowcore))
 324                 BUG();
 325 #endif
 326         vdso_init_cr5();
 327 #endif /* CONFIG_64BIT */
 328
 329         get_page(virt_to_page(vdso_data));
 330
 331         smp_wmb();
 332
 333         return 0;
 334 }
 335 arch_initcall(vdso_init);
 336
 337 int in_gate_area_no_task(unsigned long addr)
 338 {
 339         return 0;
 340 }
 341
 342 int in_gate_area(struct task_struct *task, unsigned long addr)
 343 {
 344         return 0;
 345 }
 346
 347 struct vm_area_struct *get_gate_vma(struct task_struct *tsk)
 348 {
 349         return NULL;
 350 }