ttm_bo_vm.c 11.3 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
/**************************************************************************
 *
 * Copyright (c) 2006-2009 VMware, Inc., Palo Alto, CA., USA
 * All Rights Reserved.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the
 * "Software"), to deal in the Software without restriction, including
 * without limitation the rights to use, copy, modify, merge, publish,
 * distribute, sub license, and/or sell copies of the Software, and to
 * permit persons to whom the Software is furnished to do so, subject to
 * the following conditions:
 *
 * The above copyright notice and this permission notice (including the
 * next paragraph) shall be included in all copies or substantial portions
 * of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
 * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
 * USE OR OTHER DEALINGS IN THE SOFTWARE.
 *
 **************************************************************************/
/*
 * Authors: Thomas Hellstrom <thellstrom-at-vmware-dot-com>
 */

J
Joe Perches 已提交
31 32
#define pr_fmt(fmt) "[TTM] " fmt

33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87
#include <ttm/ttm_module.h>
#include <ttm/ttm_bo_driver.h>
#include <ttm/ttm_placement.h>
#include <linux/mm.h>
#include <linux/rbtree.h>
#include <linux/module.h>
#include <linux/uaccess.h>

#define TTM_BO_VM_NUM_PREFAULT 16

static struct ttm_buffer_object *ttm_bo_vm_lookup_rb(struct ttm_bo_device *bdev,
						     unsigned long page_start,
						     unsigned long num_pages)
{
	struct rb_node *cur = bdev->addr_space_rb.rb_node;
	unsigned long cur_offset;
	struct ttm_buffer_object *bo;
	struct ttm_buffer_object *best_bo = NULL;

	while (likely(cur != NULL)) {
		bo = rb_entry(cur, struct ttm_buffer_object, vm_rb);
		cur_offset = bo->vm_node->start;
		if (page_start >= cur_offset) {
			cur = cur->rb_right;
			best_bo = bo;
			if (page_start == cur_offset)
				break;
		} else
			cur = cur->rb_left;
	}

	if (unlikely(best_bo == NULL))
		return NULL;

	if (unlikely((best_bo->vm_node->start + best_bo->num_pages) <
		     (page_start + num_pages)))
		return NULL;

	return best_bo;
}

static int ttm_bo_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
	struct ttm_buffer_object *bo = (struct ttm_buffer_object *)
	    vma->vm_private_data;
	struct ttm_bo_device *bdev = bo->bdev;
	unsigned long page_offset;
	unsigned long page_last;
	unsigned long pfn;
	struct ttm_tt *ttm = NULL;
	struct page *page;
	int ret;
	int i;
	unsigned long address = (unsigned long)vmf->virtual_address;
	int retval = VM_FAULT_NOPAGE;
88 89
	struct ttm_mem_type_manager *man =
		&bdev->man[bo->mem.mem_type];
90 91 92 93 94 95 96 97 98 99 100 101 102 103

	/*
	 * Work around locking order reversal in fault / nopfn
	 * between mmap_sem and bo_reserve: Perform a trylock operation
	 * for reserve, and if it fails, retry the fault after scheduling.
	 */

	ret = ttm_bo_reserve(bo, true, true, false, 0);
	if (unlikely(ret != 0)) {
		if (ret == -EBUSY)
			set_need_resched();
		return VM_FAULT_NOPAGE;
	}

104 105 106 107 108 109 110 111 112 113 114 115 116 117 118
	if (bdev->driver->fault_reserve_notify) {
		ret = bdev->driver->fault_reserve_notify(bo);
		switch (ret) {
		case 0:
			break;
		case -EBUSY:
			set_need_resched();
		case -ERESTARTSYS:
			retval = VM_FAULT_NOPAGE;
			goto out_unlock;
		default:
			retval = VM_FAULT_SIGBUS;
			goto out_unlock;
		}
	}
119

120 121 122 123 124
	/*
	 * Wait for buffer data in transit, due to a pipelined
	 * move.
	 */

125
	spin_lock(&bdev->fence_lock);
126 127
	if (test_bit(TTM_BO_PRIV_FLAG_MOVING, &bo->priv_flags)) {
		ret = ttm_bo_wait(bo, false, true, false);
128
		spin_unlock(&bdev->fence_lock);
129
		if (unlikely(ret != 0)) {
130
			retval = (ret != -ERESTARTSYS) ?
131 132 133 134
			    VM_FAULT_SIGBUS : VM_FAULT_NOPAGE;
			goto out_unlock;
		}
	} else
135
		spin_unlock(&bdev->fence_lock);
136

137 138 139
	ret = ttm_mem_io_lock(man, true);
	if (unlikely(ret != 0)) {
		retval = VM_FAULT_NOPAGE;
140 141
		goto out_unlock;
	}
142 143 144 145 146
	ret = ttm_mem_io_reserve_vm(bo);
	if (unlikely(ret != 0)) {
		retval = VM_FAULT_SIGBUS;
		goto out_io_unlock;
	}
147 148 149

	page_offset = ((address - vma->vm_start) >> PAGE_SHIFT) +
	    bo->vm_node->start - vma->vm_pgoff;
150
	page_last = vma_pages(vma) +
151 152 153 154
	    bo->vm_node->start - vma->vm_pgoff;

	if (unlikely(page_offset >= bo->num_pages)) {
		retval = VM_FAULT_SIGBUS;
155
		goto out_io_unlock;
156 157 158 159 160 161 162 163 164 165 166 167 168 169 170
	}

	/*
	 * Strictly, we're not allowed to modify vma->vm_page_prot here,
	 * since the mmap_sem is only held in read mode. However, we
	 * modify only the caching bits of vma->vm_page_prot and
	 * consider those bits protected by
	 * the bo->mutex, as we should be the only writers.
	 * There shouldn't really be any readers of these bits except
	 * within vm_insert_mixed()? fork?
	 *
	 * TODO: Add a list of vmas to the bo, and change the
	 * vma->vm_page_prot when the object changes caching policy, with
	 * the correct locks held.
	 */
171
	if (bo->mem.bus.is_iomem) {
172 173 174 175 176 177 178
		vma->vm_page_prot = ttm_io_prot(bo->mem.placement,
						vma->vm_page_prot);
	} else {
		ttm = bo->ttm;
		vma->vm_page_prot = (bo->mem.placement & TTM_PL_FLAG_CACHED) ?
		    vm_get_page_prot(vma->vm_flags) :
		    ttm_io_prot(bo->mem.placement, vma->vm_page_prot);
179 180 181 182 183 184

		/* Allocate all page at once, most common usage */
		if (ttm->bdev->driver->ttm_tt_populate(ttm)) {
			retval = VM_FAULT_OOM;
			goto out_io_unlock;
		}
185 186 187 188 189 190 191
	}

	/*
	 * Speculatively prefault a number of pages. Only error on
	 * first page.
	 */
	for (i = 0; i < TTM_BO_VM_NUM_PREFAULT; ++i) {
192 193
		if (bo->mem.bus.is_iomem)
			pfn = ((bo->mem.bus.base + bo->mem.bus.offset) >> PAGE_SHIFT) + page_offset;
194
		else {
195
			page = ttm->pages[page_offset];
196 197
			if (unlikely(!page && i == 0)) {
				retval = VM_FAULT_OOM;
198
				goto out_io_unlock;
199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215
			} else if (unlikely(!page)) {
				break;
			}
			pfn = page_to_pfn(page);
		}

		ret = vm_insert_mixed(vma, address, pfn);
		/*
		 * Somebody beat us to this PTE or prefaulting to
		 * an already populated PTE, or prefaulting error.
		 */

		if (unlikely((ret == -EBUSY) || (ret != 0 && i > 0)))
			break;
		else if (unlikely(ret != 0)) {
			retval =
			    (ret == -ENOMEM) ? VM_FAULT_OOM : VM_FAULT_SIGBUS;
216
			goto out_io_unlock;
217 218 219 220 221 222
		}

		address += PAGE_SIZE;
		if (unlikely(++page_offset >= page_last))
			break;
	}
223 224
out_io_unlock:
	ttm_mem_io_unlock(man);
225 226 227 228 229 230 231 232 233 234 235 236 237 238 239
out_unlock:
	ttm_bo_unreserve(bo);
	return retval;
}

static void ttm_bo_vm_open(struct vm_area_struct *vma)
{
	struct ttm_buffer_object *bo =
	    (struct ttm_buffer_object *)vma->vm_private_data;

	(void)ttm_bo_reference(bo);
}

static void ttm_bo_vm_close(struct vm_area_struct *vma)
{
240
	struct ttm_buffer_object *bo = (struct ttm_buffer_object *)vma->vm_private_data;
241 242 243 244 245

	ttm_bo_unref(&bo);
	vma->vm_private_data = NULL;
}

246
static const struct vm_operations_struct ttm_bo_vm_ops = {
247 248 249 250 251 252 253 254 255 256 257 258 259 260
	.fault = ttm_bo_vm_fault,
	.open = ttm_bo_vm_open,
	.close = ttm_bo_vm_close
};

int ttm_bo_mmap(struct file *filp, struct vm_area_struct *vma,
		struct ttm_bo_device *bdev)
{
	struct ttm_bo_driver *driver;
	struct ttm_buffer_object *bo;
	int ret;

	read_lock(&bdev->vm_lock);
	bo = ttm_bo_vm_lookup_rb(bdev, vma->vm_pgoff,
261
				 vma_pages(vma));
262 263
	if (likely(bo != NULL) && !kref_get_unless_zero(&bo->kref))
		bo = NULL;
264 265 266
	read_unlock(&bdev->vm_lock);

	if (unlikely(bo == NULL)) {
J
Joe Perches 已提交
267
		pr_err("Could not find buffer object to map\n");
268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287
		return -EINVAL;
	}

	driver = bo->bdev->driver;
	if (unlikely(!driver->verify_access)) {
		ret = -EPERM;
		goto out_unref;
	}
	ret = driver->verify_access(bo, filp);
	if (unlikely(ret != 0))
		goto out_unref;

	vma->vm_ops = &ttm_bo_vm_ops;

	/*
	 * Note: We're transferring the bo reference to
	 * vma->vm_private_data here.
	 */

	vma->vm_private_data = bo;
288
	vma->vm_flags |= VM_IO | VM_MIXEDMAP | VM_DONTEXPAND | VM_DONTDUMP;
289 290 291 292 293 294 295 296 297 298 299 300 301 302
	return 0;
out_unref:
	ttm_bo_unref(&bo);
	return ret;
}
EXPORT_SYMBOL(ttm_bo_mmap);

int ttm_fbdev_mmap(struct vm_area_struct *vma, struct ttm_buffer_object *bo)
{
	if (vma->vm_pgoff != 0)
		return -EACCES;

	vma->vm_ops = &ttm_bo_vm_ops;
	vma->vm_private_data = ttm_bo_reference(bo);
303
	vma->vm_flags |= VM_IO | VM_MIXEDMAP | VM_DONTEXPAND;
304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336
	return 0;
}
EXPORT_SYMBOL(ttm_fbdev_mmap);


ssize_t ttm_bo_io(struct ttm_bo_device *bdev, struct file *filp,
		  const char __user *wbuf, char __user *rbuf, size_t count,
		  loff_t *f_pos, bool write)
{
	struct ttm_buffer_object *bo;
	struct ttm_bo_driver *driver;
	struct ttm_bo_kmap_obj map;
	unsigned long dev_offset = (*f_pos >> PAGE_SHIFT);
	unsigned long kmap_offset;
	unsigned long kmap_end;
	unsigned long kmap_num;
	size_t io_size;
	unsigned int page_offset;
	char *virtual;
	int ret;
	bool no_wait = false;
	bool dummy;

	read_lock(&bdev->vm_lock);
	bo = ttm_bo_vm_lookup_rb(bdev, dev_offset, 1);
	if (likely(bo != NULL))
		ttm_bo_reference(bo);
	read_unlock(&bdev->vm_lock);

	if (unlikely(bo == NULL))
		return -EFAULT;

	driver = bo->bdev->driver;
337
	if (unlikely(!driver->verify_access)) {
338 339 340 341 342 343 344 345 346
		ret = -EPERM;
		goto out_unref;
	}

	ret = driver->verify_access(bo, filp);
	if (unlikely(ret != 0))
		goto out_unref;

	kmap_offset = dev_offset - bo->vm_node->start;
R
Roel Kluin 已提交
347
	if (unlikely(kmap_offset >= bo->num_pages)) {
348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417
		ret = -EFBIG;
		goto out_unref;
	}

	page_offset = *f_pos & ~PAGE_MASK;
	io_size = bo->num_pages - kmap_offset;
	io_size = (io_size << PAGE_SHIFT) - page_offset;
	if (count < io_size)
		io_size = count;

	kmap_end = (*f_pos + count - 1) >> PAGE_SHIFT;
	kmap_num = kmap_end - kmap_offset + 1;

	ret = ttm_bo_reserve(bo, true, no_wait, false, 0);

	switch (ret) {
	case 0:
		break;
	case -EBUSY:
		ret = -EAGAIN;
		goto out_unref;
	default:
		goto out_unref;
	}

	ret = ttm_bo_kmap(bo, kmap_offset, kmap_num, &map);
	if (unlikely(ret != 0)) {
		ttm_bo_unreserve(bo);
		goto out_unref;
	}

	virtual = ttm_kmap_obj_virtual(&map, &dummy);
	virtual += page_offset;

	if (write)
		ret = copy_from_user(virtual, wbuf, io_size);
	else
		ret = copy_to_user(rbuf, virtual, io_size);

	ttm_bo_kunmap(&map);
	ttm_bo_unreserve(bo);
	ttm_bo_unref(&bo);

	if (unlikely(ret != 0))
		return -EFBIG;

	*f_pos += io_size;

	return io_size;
out_unref:
	ttm_bo_unref(&bo);
	return ret;
}

ssize_t ttm_bo_fbdev_io(struct ttm_buffer_object *bo, const char __user *wbuf,
			char __user *rbuf, size_t count, loff_t *f_pos,
			bool write)
{
	struct ttm_bo_kmap_obj map;
	unsigned long kmap_offset;
	unsigned long kmap_end;
	unsigned long kmap_num;
	size_t io_size;
	unsigned int page_offset;
	char *virtual;
	int ret;
	bool no_wait = false;
	bool dummy;

	kmap_offset = (*f_pos >> PAGE_SHIFT);
R
Roel Kluin 已提交
418
	if (unlikely(kmap_offset >= bo->num_pages))
419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465
		return -EFBIG;

	page_offset = *f_pos & ~PAGE_MASK;
	io_size = bo->num_pages - kmap_offset;
	io_size = (io_size << PAGE_SHIFT) - page_offset;
	if (count < io_size)
		io_size = count;

	kmap_end = (*f_pos + count - 1) >> PAGE_SHIFT;
	kmap_num = kmap_end - kmap_offset + 1;

	ret = ttm_bo_reserve(bo, true, no_wait, false, 0);

	switch (ret) {
	case 0:
		break;
	case -EBUSY:
		return -EAGAIN;
	default:
		return ret;
	}

	ret = ttm_bo_kmap(bo, kmap_offset, kmap_num, &map);
	if (unlikely(ret != 0)) {
		ttm_bo_unreserve(bo);
		return ret;
	}

	virtual = ttm_kmap_obj_virtual(&map, &dummy);
	virtual += page_offset;

	if (write)
		ret = copy_from_user(virtual, wbuf, io_size);
	else
		ret = copy_to_user(rbuf, virtual, io_size);

	ttm_bo_kunmap(&map);
	ttm_bo_unreserve(bo);
	ttm_bo_unref(&bo);

	if (unlikely(ret != 0))
		return ret;

	*f_pos += io_size;

	return io_size;
}