drivers/dax/device.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 /* Copyright(c) 2016-2018 Intel Corporation. All rights reserved. */
   3 #include <linux/memremap.h>
   4 #include <linux/pagemap.h>
   5 #include <linux/module.h>
   6 #include <linux/device.h>
   7 #include <linux/pfn_t.h>
   8 #include <linux/cdev.h>
   9 #include <linux/slab.h>
  10 #include <linux/dax.h>
  11 #include <linux/fs.h>
  12 #include <linux/mm.h>
  13 #include <linux/mman.h>
  14 #include "dax-private.h"
  15 #include "bus.h"
  16
  17 static int check_vma(struct dev_dax *dev_dax, struct vm_area_struct *vma,
  18                 const char *func)
  19 {
  20         struct dax_region *dax_region = dev_dax->region;
  21         struct device *dev = &dev_dax->dev;
  22         unsigned long mask;
  23
  24         if (!dax_alive(dev_dax->dax_dev))
  25                 return -ENXIO;
  26
  27         /* prevent private mappings from being established */
  28         if ((vma->vm_flags & VM_MAYSHARE) != VM_MAYSHARE) {
  29                 dev_info_ratelimited(dev,
  30                                 "%s: %s: fail, attempted private mapping\n",
  31                                 current->comm, func);
  32                 return -EINVAL;
  33         }
  34
  35         mask = dax_region->align - 1;
  36         if (vma->vm_start & mask || vma->vm_end & mask) {
  37                 dev_info_ratelimited(dev,
  38                                 "%s: %s: fail, unaligned vma (%#lx - %#lx, %#lx)\n",
  39                                 current->comm, func, vma->vm_start, vma->vm_end,
  40                                 mask);
  41                 return -EINVAL;
  42         }
  43
  44         if (!vma_is_dax(vma)) {
  45                 dev_info_ratelimited(dev,
  46                                 "%s: %s: fail, vma is not DAX capable\n",
  47                                 current->comm, func);
  48                 return -EINVAL;
  49         }
  50
  51         return 0;
  52 }
  53
  54 /* see "strong" declaration in tools/testing/nvdimm/dax-dev.c */
  55 __weak phys_addr_t dax_pgoff_to_phys(struct dev_dax *dev_dax, pgoff_t pgoff,
  56                 unsigned long size)
  57 {
  58         struct range *range = &dev_dax->range;
  59         phys_addr_t phys;
  60
  61         phys = pgoff * PAGE_SIZE + range->start;
  62         if (phys >= range->start && phys <= range->end) {
  63                 if (phys + size - 1 <= range->end)
  64                         return phys;
  65         }
  66
  67         return -1;
  68 }
  69
  70 static vm_fault_t __dev_dax_pte_fault(struct dev_dax *dev_dax,
  71                                 struct vm_fault *vmf, pfn_t *pfn)
  72 {
  73         struct device *dev = &dev_dax->dev;
  74         struct dax_region *dax_region;
  75         phys_addr_t phys;
  76         unsigned int fault_size = PAGE_SIZE;
  77
  78         if (check_vma(dev_dax, vmf->vma, __func__))
  79                 return VM_FAULT_SIGBUS;
  80
  81         dax_region = dev_dax->region;
  82         if (dax_region->align > PAGE_SIZE) {
  83                 dev_dbg(dev, "alignment (%#x) > fault size (%#x)\n",
  84                         dax_region->align, fault_size);
  85                 return VM_FAULT_SIGBUS;
  86         }
  87
  88         if (fault_size != dax_region->align)
  89                 return VM_FAULT_SIGBUS;
  90
  91         phys = dax_pgoff_to_phys(dev_dax, vmf->pgoff, PAGE_SIZE);
  92         if (phys == -1) {
  93                 dev_dbg(dev, "pgoff_to_phys(%#lx) failed\n", vmf->pgoff);
  94                 return VM_FAULT_SIGBUS;
  95         }
  96
  97         *pfn = phys_to_pfn_t(phys, PFN_DEV|PFN_MAP);
  98
  99         return vmf_insert_mixed(vmf->vma, vmf->address, *pfn);
 100 }
 101
 102 static vm_fault_t __dev_dax_pmd_fault(struct dev_dax *dev_dax,
 103                                 struct vm_fault *vmf, pfn_t *pfn)
 104 {
 105         unsigned long pmd_addr = vmf->address & PMD_MASK;
 106         struct device *dev = &dev_dax->dev;
 107         struct dax_region *dax_region;
 108         phys_addr_t phys;
 109         pgoff_t pgoff;
 110         unsigned int fault_size = PMD_SIZE;
 111
 112         if (check_vma(dev_dax, vmf->vma, __func__))
 113                 return VM_FAULT_SIGBUS;
 114
 115         dax_region = dev_dax->region;
 116         if (dax_region->align > PMD_SIZE) {
 117                 dev_dbg(dev, "alignment (%#x) > fault size (%#x)\n",
 118                         dax_region->align, fault_size);
 119                 return VM_FAULT_SIGBUS;
 120         }
 121
 122         if (fault_size < dax_region->align)
 123                 return VM_FAULT_SIGBUS;
 124         else if (fault_size > dax_region->align)
 125                 return VM_FAULT_FALLBACK;
 126
 127         /* if we are outside of the VMA */
 128         if (pmd_addr < vmf->vma->vm_start ||
 129                         (pmd_addr + PMD_SIZE) > vmf->vma->vm_end)
 130                 return VM_FAULT_SIGBUS;
 131
 132         pgoff = linear_page_index(vmf->vma, pmd_addr);
 133         phys = dax_pgoff_to_phys(dev_dax, pgoff, PMD_SIZE);
 134         if (phys == -1) {
 135                 dev_dbg(dev, "pgoff_to_phys(%#lx) failed\n", pgoff);
 136                 return VM_FAULT_SIGBUS;
 137         }
 138
 139         *pfn = phys_to_pfn_t(phys, PFN_DEV|PFN_MAP);
 140
 141         return vmf_insert_pfn_pmd(vmf, *pfn, vmf->flags & FAULT_FLAG_WRITE);
 142 }
 143
 144 #ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
 145 static vm_fault_t __dev_dax_pud_fault(struct dev_dax *dev_dax,
 146                                 struct vm_fault *vmf, pfn_t *pfn)
 147 {
 148         unsigned long pud_addr = vmf->address & PUD_MASK;
 149         struct device *dev = &dev_dax->dev;
 150         struct dax_region *dax_region;
 151         phys_addr_t phys;
 152         pgoff_t pgoff;
 153         unsigned int fault_size = PUD_SIZE;
 154
 155
 156         if (check_vma(dev_dax, vmf->vma, __func__))
 157                 return VM_FAULT_SIGBUS;
 158
 159         dax_region = dev_dax->region;
 160         if (dax_region->align > PUD_SIZE) {
 161                 dev_dbg(dev, "alignment (%#x) > fault size (%#x)\n",
 162                         dax_region->align, fault_size);
 163                 return VM_FAULT_SIGBUS;
 164         }
 165
 166         if (fault_size < dax_region->align)
 167                 return VM_FAULT_SIGBUS;
 168         else if (fault_size > dax_region->align)
 169                 return VM_FAULT_FALLBACK;
 170
 171         /* if we are outside of the VMA */
 172         if (pud_addr < vmf->vma->vm_start ||
 173                         (pud_addr + PUD_SIZE) > vmf->vma->vm_end)
 174                 return VM_FAULT_SIGBUS;
 175
 176         pgoff = linear_page_index(vmf->vma, pud_addr);
 177         phys = dax_pgoff_to_phys(dev_dax, pgoff, PUD_SIZE);
 178         if (phys == -1) {
 179                 dev_dbg(dev, "pgoff_to_phys(%#lx) failed\n", pgoff);
 180                 return VM_FAULT_SIGBUS;
 181         }
 182
 183         *pfn = phys_to_pfn_t(phys, PFN_DEV|PFN_MAP);
 184
 185         return vmf_insert_pfn_pud(vmf, *pfn, vmf->flags & FAULT_FLAG_WRITE);
 186 }
 187 #else
 188 static vm_fault_t __dev_dax_pud_fault(struct dev_dax *dev_dax,
 189                                 struct vm_fault *vmf, pfn_t *pfn)
 190 {
 191         return VM_FAULT_FALLBACK;
 192 }
 193 #endif /* !CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD */
 194
 195 static vm_fault_t dev_dax_huge_fault(struct vm_fault *vmf,
 196                 enum page_entry_size pe_size)
 197 {
 198         struct file *filp = vmf->vma->vm_file;
 199         unsigned long fault_size;
 200         vm_fault_t rc = VM_FAULT_SIGBUS;
 201         int id;
 202         pfn_t pfn;
 203         struct dev_dax *dev_dax = filp->private_data;
 204
 205         dev_dbg(&dev_dax->dev, "%s: %s (%#lx - %#lx) size = %d\n", current->comm,
 206                         (vmf->flags & FAULT_FLAG_WRITE) ? "write" : "read",
 207                         vmf->vma->vm_start, vmf->vma->vm_end, pe_size);
 208
 209         id = dax_read_lock();
 210         switch (pe_size) {
 211         case PE_SIZE_PTE:
 212                 fault_size = PAGE_SIZE;
 213                 rc = __dev_dax_pte_fault(dev_dax, vmf, &pfn);
 214                 break;
 215         case PE_SIZE_PMD:
 216                 fault_size = PMD_SIZE;
 217                 rc = __dev_dax_pmd_fault(dev_dax, vmf, &pfn);
 218                 break;
 219         case PE_SIZE_PUD:
 220                 fault_size = PUD_SIZE;
 221                 rc = __dev_dax_pud_fault(dev_dax, vmf, &pfn);
 222                 break;
 223         default:
 224                 rc = VM_FAULT_SIGBUS;
 225         }
 226
 227         if (rc == VM_FAULT_NOPAGE) {
 228                 unsigned long i;
 229                 pgoff_t pgoff;
 230
 231                 /*
 232                  * In the device-dax case the only possibility for a
 233                  * VM_FAULT_NOPAGE result is when device-dax capacity is
 234                  * mapped. No need to consider the zero page, or racing
 235                  * conflicting mappings.
 236                  */
 237                 pgoff = linear_page_index(vmf->vma, vmf->address
 238                                 & ~(fault_size - 1));
 239                 for (i = 0; i < fault_size / PAGE_SIZE; i++) {
 240                         struct page *page;
 241
 242                         page = pfn_to_page(pfn_t_to_pfn(pfn) + i);
 243                         if (page->mapping)
 244                                 continue;
 245                         page->mapping = filp->f_mapping;
 246                         page->index = pgoff + i;
 247                 }
 248         }
 249         dax_read_unlock(id);
 250
 251         return rc;
 252 }
 253
 254 static vm_fault_t dev_dax_fault(struct vm_fault *vmf)
 255 {
 256         return dev_dax_huge_fault(vmf, PE_SIZE_PTE);
 257 }
 258
 259 static int dev_dax_split(struct vm_area_struct *vma, unsigned long addr)
 260 {
 261         struct file *filp = vma->vm_file;
 262         struct dev_dax *dev_dax = filp->private_data;
 263         struct dax_region *dax_region = dev_dax->region;
 264
 265         if (!IS_ALIGNED(addr, dax_region->align))
 266                 return -EINVAL;
 267         return 0;
 268 }
 269
 270 static unsigned long dev_dax_pagesize(struct vm_area_struct *vma)
 271 {
 272         struct file *filp = vma->vm_file;
 273         struct dev_dax *dev_dax = filp->private_data;
 274         struct dax_region *dax_region = dev_dax->region;
 275
 276         return dax_region->align;
 277 }
 278
 279 static const struct vm_operations_struct dax_vm_ops = {
 280         .fault = dev_dax_fault,
 281         .huge_fault = dev_dax_huge_fault,
 282         .split = dev_dax_split,
 283         .pagesize = dev_dax_pagesize,
 284 };
 285
 286 static int dax_mmap(struct file *filp, struct vm_area_struct *vma)
 287 {
 288         struct dev_dax *dev_dax = filp->private_data;
 289         int rc, id;
 290
 291         dev_dbg(&dev_dax->dev, "trace\n");
 292
 293         /*
 294          * We lock to check dax_dev liveness and will re-check at
 295          * fault time.
 296          */
 297         id = dax_read_lock();
 298         rc = check_vma(dev_dax, vma, __func__);
 299         dax_read_unlock(id);
 300         if (rc)
 301                 return rc;
 302
 303         vma->vm_ops = &dax_vm_ops;
 304         vma->vm_flags |= VM_HUGEPAGE;
 305         return 0;
 306 }
 307
 308 /* return an unmapped area aligned to the dax region specified alignment */
 309 static unsigned long dax_get_unmapped_area(struct file *filp,
 310                 unsigned long addr, unsigned long len, unsigned long pgoff,
 311                 unsigned long flags)
 312 {
 313         unsigned long off, off_end, off_align, len_align, addr_align, align;
 314         struct dev_dax *dev_dax = filp ? filp->private_data : NULL;
 315         struct dax_region *dax_region;
 316
 317         if (!dev_dax || addr)
 318                 goto out;
 319
 320         dax_region = dev_dax->region;
 321         align = dax_region->align;
 322         off = pgoff << PAGE_SHIFT;
 323         off_end = off + len;
 324         off_align = round_up(off, align);
 325
 326         if ((off_end <= off_align) || ((off_end - off_align) < align))
 327                 goto out;
 328
 329         len_align = len + align;
 330         if ((off + len_align) < off)
 331                 goto out;
 332
 333         addr_align = current->mm->get_unmapped_area(filp, addr, len_align,
 334                         pgoff, flags);
 335         if (!IS_ERR_VALUE(addr_align)) {
 336                 addr_align += (off - addr_align) & (align - 1);
 337                 return addr_align;
 338         }
 339  out:
 340         return current->mm->get_unmapped_area(filp, addr, len, pgoff, flags);
 341 }
 342
 343 static const struct address_space_operations dev_dax_aops = {
 344         .set_page_dirty         = noop_set_page_dirty,
 345         .invalidatepage         = noop_invalidatepage,
 346 };
 347
 348 static int dax_open(struct inode *inode, struct file *filp)
 349 {
 350         struct dax_device *dax_dev = inode_dax(inode);
 351         struct inode *__dax_inode = dax_inode(dax_dev);
 352         struct dev_dax *dev_dax = dax_get_private(dax_dev);
 353
 354         dev_dbg(&dev_dax->dev, "trace\n");
 355         inode->i_mapping = __dax_inode->i_mapping;
 356         inode->i_mapping->host = __dax_inode;
 357         inode->i_mapping->a_ops = &dev_dax_aops;
 358         filp->f_mapping = inode->i_mapping;
 359         filp->f_wb_err = filemap_sample_wb_err(filp->f_mapping);
 360         filp->f_sb_err = file_sample_sb_err(filp);
 361         filp->private_data = dev_dax;
 362         inode->i_flags = S_DAX;
 363
 364         return 0;
 365 }
 366
 367 static int dax_release(struct inode *inode, struct file *filp)
 368 {
 369         struct dev_dax *dev_dax = filp->private_data;
 370
 371         dev_dbg(&dev_dax->dev, "trace\n");
 372         return 0;
 373 }
 374
 375 static const struct file_operations dax_fops = {
 376         .llseek = noop_llseek,
 377         .owner = THIS_MODULE,
 378         .open = dax_open,
 379         .release = dax_release,
 380         .get_unmapped_area = dax_get_unmapped_area,
 381         .mmap = dax_mmap,
 382         .mmap_supported_flags = MAP_SYNC,
 383 };
 384
 385 static void dev_dax_cdev_del(void *cdev)
 386 {
 387         cdev_del(cdev);
 388 }
 389
 390 static void dev_dax_kill(void *dev_dax)
 391 {
 392         kill_dev_dax(dev_dax);
 393 }
 394
 395 int dev_dax_probe(struct dev_dax *dev_dax)
 396 {
 397         struct dax_device *dax_dev = dev_dax->dax_dev;
 398         struct range *range = &dev_dax->range;
 399         struct device *dev = &dev_dax->dev;
 400         struct dev_pagemap *pgmap;
 401         struct inode *inode;
 402         struct cdev *cdev;
 403         void *addr;
 404         int rc;
 405
 406         /* 1:1 map region resource range to device-dax instance range */
 407         if (!devm_request_mem_region(dev, range->start, range_len(range),
 408                                 dev_name(dev))) {
 409                 dev_warn(dev, "could not reserve range: %#llx - %#llx\n",
 410                                 range->start, range->end);
 411                 return -EBUSY;
 412         }
 413
 414         pgmap = dev_dax->pgmap;
 415         if (!pgmap) {
 416                 pgmap = devm_kzalloc(dev, sizeof(*pgmap), GFP_KERNEL);
 417                 if (!pgmap)
 418                         return -ENOMEM;
 419                 pgmap->range = *range;
 420         }
 421         pgmap->type = MEMORY_DEVICE_GENERIC;
 422         addr = devm_memremap_pages(dev, pgmap);
 423         if (IS_ERR(addr))
 424                 return PTR_ERR(addr);
 425
 426         inode = dax_inode(dax_dev);
 427         cdev = inode->i_cdev;
 428         cdev_init(cdev, &dax_fops);
 429         if (dev->class) {
 430                 /* for the CONFIG_DEV_DAX_PMEM_COMPAT case */
 431                 cdev->owner = dev->parent->driver->owner;
 432         } else
 433                 cdev->owner = dev->driver->owner;
 434         cdev_set_parent(cdev, &dev->kobj);
 435         rc = cdev_add(cdev, dev->devt, 1);
 436         if (rc)
 437                 return rc;
 438
 439         rc = devm_add_action_or_reset(dev, dev_dax_cdev_del, cdev);
 440         if (rc)
 441                 return rc;
 442
 443         run_dax(dax_dev);
 444         return devm_add_action_or_reset(dev, dev_dax_kill, dev_dax);
 445 }
 446 EXPORT_SYMBOL_GPL(dev_dax_probe);
 447
 448 static int dev_dax_remove(struct dev_dax *dev_dax)
 449 {
 450         /* all probe actions are unwound by devm */
 451         return 0;
 452 }
 453
 454 static struct dax_device_driver device_dax_driver = {
 455         .probe = dev_dax_probe,
 456         .remove = dev_dax_remove,
 457         .match_always = 1,
 458 };
 459
 460 static int __init dax_init(void)
 461 {
 462         return dax_driver_register(&device_dax_driver);
 463 }
 464
 465 static void __exit dax_exit(void)
 466 {
 467         dax_driver_unregister(&device_dax_driver);
 468 }
 469
 470 MODULE_AUTHOR("Intel Corporation");
 471 MODULE_LICENSE("GPL v2");
 472 module_init(dax_init);
 473 module_exit(dax_exit);
 474 MODULE_ALIAS_DAX_DEVICE(0);