arch/arm64/kernel/vdso.c

   1 /*
   2  * VDSO implementation for AArch64 and vector page setup for AArch32.
   3  *
   4  * Copyright (C) 2012 ARM Limited
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License version 2 as
   8  * published by the Free Software Foundation.
   9  *
  10  * This program is distributed in the hope that it will be useful,
  11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  13  * GNU General Public License for more details.
  14  *
  15  * You should have received a copy of the GNU General Public License
  16  * along with this program.  If not, see <http://www.gnu.org/licenses/>.
  17  *
  18  * Author: Will Deacon <will.deacon@arm.com>
  19  */
  20
  21 #include <linux/cache.h>
  22 #include <linux/clocksource.h>
  23 #include <linux/elf.h>
  24 #include <linux/err.h>
  25 #include <linux/errno.h>
  26 #include <linux/gfp.h>
  27 #include <linux/kernel.h>
  28 #include <linux/mm.h>
  29 #include <linux/sched.h>
  30 #include <linux/signal.h>
  31 #include <linux/slab.h>
  32 #include <linux/timekeeper_internal.h>
  33 #include <linux/vmalloc.h>
  34
  35 #include <asm/cacheflush.h>
  36 #include <asm/signal32.h>
  37 #include <asm/vdso.h>
  38 #include <asm/vdso_datapage.h>
  39
  40 extern char vdso_start, vdso_end;
  41 static unsigned long vdso_pages __ro_after_init;
  42
  43 /*
  44  * The vDSO data page.
  45  */
  46 static union {
  47         struct vdso_data        data;
  48         u8                      page[PAGE_SIZE];
  49 } vdso_data_store __page_aligned_data;
  50 struct vdso_data *vdso_data = &vdso_data_store.data;
  51
  52 #ifdef CONFIG_COMPAT
  53 /*
  54  * Create and map the vectors page for AArch32 tasks.
  55  */
  56 static struct page *vectors_page[1] __ro_after_init;
  57
  58 static int __init alloc_vectors_page(void)
  59 {
  60         extern char __kuser_helper_start[], __kuser_helper_end[];
  61         extern char __aarch32_sigret_code_start[], __aarch32_sigret_code_end[];
  62
  63         int kuser_sz = __kuser_helper_end - __kuser_helper_start;
  64         int sigret_sz = __aarch32_sigret_code_end - __aarch32_sigret_code_start;
  65         unsigned long vpage;
  66
  67         vpage = get_zeroed_page(GFP_ATOMIC);
  68
  69         if (!vpage)
  70                 return -ENOMEM;
  71
  72         /* kuser helpers */
  73         memcpy((void *)vpage + 0x1000 - kuser_sz, __kuser_helper_start,
  74                 kuser_sz);
  75
  76         /* sigreturn code */
  77         memcpy((void *)vpage + AARCH32_KERN_SIGRET_CODE_OFFSET,
  78                __aarch32_sigret_code_start, sigret_sz);
  79
  80         flush_icache_range(vpage, vpage + PAGE_SIZE);
  81         vectors_page[0] = virt_to_page(vpage);
  82
  83         return 0;
  84 }
  85 arch_initcall(alloc_vectors_page);
  86
  87 int aarch32_setup_vectors_page(struct linux_binprm *bprm, int uses_interp)
  88 {
  89         struct mm_struct *mm = current->mm;
  90         unsigned long addr = AARCH32_VECTORS_BASE;
  91         static const struct vm_special_mapping spec = {
  92                 .name   = "[vectors]",
  93                 .pages  = vectors_page,
  94
  95         };
  96         void *ret;
  97
  98         if (down_write_killable(&mm->mmap_sem))
  99                 return -EINTR;
 100         current->mm->context.vdso = (void *)addr;
 101
 102         /* Map vectors page at the high address. */
 103         ret = _install_special_mapping(mm, addr, PAGE_SIZE,
 104                                        VM_READ|VM_EXEC|VM_MAYREAD|VM_MAYEXEC,
 105                                        &spec);
 106
 107         up_write(&mm->mmap_sem);
 108
 109         return PTR_ERR_OR_ZERO(ret);
 110 }
 111 #endif /* CONFIG_COMPAT */
 112
 113 static struct vm_special_mapping vdso_spec[2] __ro_after_init = {
 114         {
 115                 .name   = "[vvar]",
 116         },
 117         {
 118                 .name   = "[vdso]",
 119         },
 120 };
 121
 122 static int __init vdso_init(void)
 123 {
 124         int i;
 125         struct page **vdso_pagelist;
 126
 127         if (memcmp(&vdso_start, "\177ELF", 4)) {
 128                 pr_err("vDSO is not a valid ELF object!\n");
 129                 return -EINVAL;
 130         }
 131
 132         vdso_pages = (&vdso_end - &vdso_start) >> PAGE_SHIFT;
 133         pr_info("vdso: %ld pages (%ld code @ %p, %ld data @ %p)\n",
 134                 vdso_pages + 1, vdso_pages, &vdso_start, 1L, vdso_data);
 135
 136         /* Allocate the vDSO pagelist, plus a page for the data. */
 137         vdso_pagelist = kcalloc(vdso_pages + 1, sizeof(struct page *),
 138                                 GFP_KERNEL);
 139         if (vdso_pagelist == NULL)
 140                 return -ENOMEM;
 141
 142         /* Grab the vDSO data page. */
 143         vdso_pagelist[0] = pfn_to_page(PHYS_PFN(__pa(vdso_data)));
 144
 145         /* Grab the vDSO code pages. */
 146         for (i = 0; i < vdso_pages; i++)
 147                 vdso_pagelist[i + 1] = pfn_to_page(PHYS_PFN(__pa(&vdso_start)) + i);
 148
 149         vdso_spec[0].pages = &vdso_pagelist[0];
 150         vdso_spec[1].pages = &vdso_pagelist[1];
 151
 152         return 0;
 153 }
 154 arch_initcall(vdso_init);
 155
 156 int arch_setup_additional_pages(struct linux_binprm *bprm,
 157                                 int uses_interp)
 158 {
 159         struct mm_struct *mm = current->mm;
 160         unsigned long vdso_base, vdso_text_len, vdso_mapping_len;
 161         void *ret;
 162
 163         vdso_text_len = vdso_pages << PAGE_SHIFT;
 164         /* Be sure to map the data page */
 165         vdso_mapping_len = vdso_text_len + PAGE_SIZE;
 166
 167         if (down_write_killable(&mm->mmap_sem))
 168                 return -EINTR;
 169         vdso_base = get_unmapped_area(NULL, 0, vdso_mapping_len, 0, 0);
 170         if (IS_ERR_VALUE(vdso_base)) {
 171                 ret = ERR_PTR(vdso_base);
 172                 goto up_fail;
 173         }
 174         ret = _install_special_mapping(mm, vdso_base, PAGE_SIZE,
 175                                        VM_READ|VM_MAYREAD,
 176                                        &vdso_spec[0]);
 177         if (IS_ERR(ret))
 178                 goto up_fail;
 179
 180         vdso_base += PAGE_SIZE;
 181         mm->context.vdso = (void *)vdso_base;
 182         ret = _install_special_mapping(mm, vdso_base, vdso_text_len,
 183                                        VM_READ|VM_EXEC|
 184                                        VM_MAYREAD|VM_MAYWRITE|VM_MAYEXEC,
 185                                        &vdso_spec[1]);
 186         if (IS_ERR(ret))
 187                 goto up_fail;
 188
 189
 190         up_write(&mm->mmap_sem);
 191         return 0;
 192
 193 up_fail:
 194         mm->context.vdso = NULL;
 195         up_write(&mm->mmap_sem);
 196         return PTR_ERR(ret);
 197 }
 198
 199 /*
 200  * Update the vDSO data page to keep in sync with kernel timekeeping.
 201  */
 202 void update_vsyscall(struct timekeeper *tk)
 203 {
 204         u32 use_syscall = strcmp(tk->tkr_mono.clock->name, "arch_sys_counter");
 205
 206         ++vdso_data->tb_seq_count;
 207         smp_wmb();
 208
 209         vdso_data->use_syscall                  = use_syscall;
 210         vdso_data->xtime_coarse_sec             = tk->xtime_sec;
 211         vdso_data->xtime_coarse_nsec            = tk->tkr_mono.xtime_nsec >>
 212                                                         tk->tkr_mono.shift;
 213         vdso_data->wtm_clock_sec                = tk->wall_to_monotonic.tv_sec;
 214         vdso_data->wtm_clock_nsec               = tk->wall_to_monotonic.tv_nsec;
 215
 216         if (!use_syscall) {
 217                 /* tkr_mono.cycle_last == tkr_raw.cycle_last */
 218                 vdso_data->cs_cycle_last        = tk->tkr_mono.cycle_last;
 219                 vdso_data->raw_time_sec         = tk->raw_time.tv_sec;
 220                 vdso_data->raw_time_nsec        = tk->raw_time.tv_nsec;
 221                 vdso_data->xtime_clock_sec      = tk->xtime_sec;
 222                 vdso_data->xtime_clock_nsec     = tk->tkr_mono.xtime_nsec;
 223                 /* tkr_raw.xtime_nsec == 0 */
 224                 vdso_data->cs_mono_mult         = tk->tkr_mono.mult;
 225                 vdso_data->cs_raw_mult          = tk->tkr_raw.mult;
 226                 /* tkr_mono.shift == tkr_raw.shift */
 227                 vdso_data->cs_shift             = tk->tkr_mono.shift;
 228         }
 229
 230         smp_wmb();
 231         ++vdso_data->tb_seq_count;
 232 }
 233
 234 void update_vsyscall_tz(void)
 235 {
 236         vdso_data->tz_minuteswest       = sys_tz.tz_minuteswest;
 237         vdso_data->tz_dsttime           = sys_tz.tz_dsttime;
 238 }