umem.c 7.6 KB
Newer Older
1 2 3
/*
 * Copyright (c) 2005 Topspin Communications.  All rights reserved.
 * Copyright (c) 2005 Cisco Systems.  All rights reserved.
4
 * Copyright (c) 2005 Mellanox Technologies. All rights reserved.
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 31 32 33 34 35 36 37 38
 *
 * This software is available to you under a choice of one of two
 * licenses.  You may choose to be licensed under the terms of the GNU
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the
 * OpenIB.org BSD license below:
 *
 *     Redistribution and use in source and binary forms, with or
 *     without modification, are permitted provided that the following
 *     conditions are met:
 *
 *      - Redistributions of source code must retain the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer.
 *
 *      - Redistributions in binary form must reproduce the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer in the documentation and/or other materials
 *        provided with the distribution.
 *
 * 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
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * 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.
 *
 * $Id: uverbs_mem.c 2743 2005-06-28 22:27:59Z roland $
 */

#include <linux/mm.h>
#include <linux/dma-mapping.h>
A
Alexey Dobriyan 已提交
39
#include <linux/sched.h>
40
#include <linux/hugetlb.h>
41 42 43

#include "uverbs.h"

44 45 46 47 48
#define IB_UMEM_MAX_PAGE_CHUNK						\
	((PAGE_SIZE - offsetof(struct ib_umem_chunk, page_list)) /	\
	 ((void *) &((struct ib_umem_chunk *) 0)->page_list[1] -	\
	  (void *) &((struct ib_umem_chunk *) 0)->page_list[0]))

49 50 51 52 53 54
static void __ib_umem_release(struct ib_device *dev, struct ib_umem *umem, int dirty)
{
	struct ib_umem_chunk *chunk, *tmp;
	int i;

	list_for_each_entry_safe(chunk, tmp, &umem->chunk_list, list) {
55 56
		ib_dma_unmap_sg(dev, chunk->page_list,
				chunk->nents, DMA_BIDIRECTIONAL);
57 58 59 60 61 62 63 64 65 66
		for (i = 0; i < chunk->nents; ++i) {
			if (umem->writable && dirty)
				set_page_dirty_lock(chunk->page_list[i].page);
			put_page(chunk->page_list[i].page);
		}

		kfree(chunk);
	}
}

67 68 69 70 71 72 73 74 75
/**
 * ib_umem_get - Pin and DMA map userspace memory.
 * @context: userspace context to pin memory for
 * @addr: userspace virtual address to start at
 * @size: length of region to pin
 * @access: IB_ACCESS_xxx flags for memory being pinned
 */
struct ib_umem *ib_umem_get(struct ib_ucontext *context, unsigned long addr,
			    size_t size, int access)
76
{
77
	struct ib_umem *umem;
78
	struct page **page_list;
79
	struct vm_area_struct **vma_list;
80 81 82 83 84
	struct ib_umem_chunk *chunk;
	unsigned long locked;
	unsigned long lock_limit;
	unsigned long cur_base;
	unsigned long npages;
85
	int ret;
86 87 88 89
	int off;
	int i;

	if (!can_do_mlock())
90
		return ERR_PTR(-EPERM);
91

92 93 94
	umem = kmalloc(sizeof *umem, GFP_KERNEL);
	if (!umem)
		return ERR_PTR(-ENOMEM);
95

96 97 98 99 100 101 102 103 104 105 106 107
	umem->context   = context;
	umem->length    = size;
	umem->offset    = addr & ~PAGE_MASK;
	umem->page_size = PAGE_SIZE;
	/*
	 * We ask for writable memory if any access flags other than
	 * "remote read" are set.  "Local write" and "remote write"
	 * obviously require write access.  "Remote atomic" can do
	 * things like fetch and add, which will modify memory, and
	 * "MW bind" can change permissions by binding a window.
	 */
	umem->writable  = !!(access & ~IB_ACCESS_REMOTE_READ);
108

109 110 111
	/* We assume the memory is from hugetlb until proved otherwise */
	umem->hugetlb   = 1;

112 113 114 115 116 117 118
	INIT_LIST_HEAD(&umem->chunk_list);

	page_list = (struct page **) __get_free_page(GFP_KERNEL);
	if (!page_list) {
		kfree(umem);
		return ERR_PTR(-ENOMEM);
	}
119

120 121 122 123 124 125 126 127
	/*
	 * if we can't alloc the vma_list, it's not so bad;
	 * just assume the memory is not hugetlb memory
	 */
	vma_list = (struct vm_area_struct **) __get_free_page(GFP_KERNEL);
	if (!vma_list)
		umem->hugetlb = 0;

128
	npages = PAGE_ALIGN(size + umem->offset) >> PAGE_SHIFT;
129 130 131 132 133 134 135 136 137 138 139

	down_write(&current->mm->mmap_sem);

	locked     = npages + current->mm->locked_vm;
	lock_limit = current->signal->rlim[RLIMIT_MEMLOCK].rlim_cur >> PAGE_SHIFT;

	if ((locked > lock_limit) && !capable(CAP_IPC_LOCK)) {
		ret = -ENOMEM;
		goto out;
	}

140
	cur_base = addr & PAGE_MASK;
141

142
	ret = 0;
143 144 145 146
	while (npages) {
		ret = get_user_pages(current, current->mm, cur_base,
				     min_t(int, npages,
					   PAGE_SIZE / sizeof (struct page *)),
147
				     1, !umem->writable, page_list, vma_list);
148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167

		if (ret < 0)
			goto out;

		cur_base += ret * PAGE_SIZE;
		npages   -= ret;

		off = 0;

		while (ret) {
			chunk = kmalloc(sizeof *chunk + sizeof (struct scatterlist) *
					min_t(int, ret, IB_UMEM_MAX_PAGE_CHUNK),
					GFP_KERNEL);
			if (!chunk) {
				ret = -ENOMEM;
				goto out;
			}

			chunk->nents = min_t(int, ret, IB_UMEM_MAX_PAGE_CHUNK);
			for (i = 0; i < chunk->nents; ++i) {
168 169 170
				if (vma_list &&
				    !is_vm_hugetlb_page(vma_list[i + off]))
					umem->hugetlb = 0;
171 172 173 174 175
				chunk->page_list[i].page   = page_list[i + off];
				chunk->page_list[i].offset = 0;
				chunk->page_list[i].length = PAGE_SIZE;
			}

176
			chunk->nmap = ib_dma_map_sg(context->device,
177 178 179
						    &chunk->page_list[0],
						    chunk->nents,
						    DMA_BIDIRECTIONAL);
180 181 182 183 184 185 186 187 188 189 190
			if (chunk->nmap <= 0) {
				for (i = 0; i < chunk->nents; ++i)
					put_page(chunk->page_list[i].page);
				kfree(chunk);

				ret = -ENOMEM;
				goto out;
			}

			ret -= chunk->nents;
			off += chunk->nents;
191
			list_add_tail(&chunk->list, &umem->chunk_list);
192 193 194 195 196 197
		}

		ret = 0;
	}

out:
198 199 200 201
	if (ret < 0) {
		__ib_umem_release(context->device, umem, 0);
		kfree(umem);
	} else
202 203 204
		current->mm->locked_vm = locked;

	up_write(&current->mm->mmap_sem);
205 206
	if (vma_list)
		free_page((unsigned long) vma_list);
207 208
	free_page((unsigned long) page_list);

209
	return ret < 0 ? ERR_PTR(ret) : umem;
210
}
211
EXPORT_SYMBOL(ib_umem_get);
212

213
static void ib_umem_account(struct work_struct *work)
214
{
215 216 217 218 219 220 221
	struct ib_umem *umem = container_of(work, struct ib_umem, work);

	down_write(&umem->mm->mmap_sem);
	umem->mm->locked_vm -= umem->diff;
	up_write(&umem->mm->mmap_sem);
	mmput(umem->mm);
	kfree(umem);
222 223
}

224 225 226 227 228
/**
 * ib_umem_release - release memory pinned with ib_umem_get
 * @umem: umem struct to release
 */
void ib_umem_release(struct ib_umem *umem)
229
{
230
	struct ib_ucontext *context = umem->context;
231
	struct mm_struct *mm;
232
	unsigned long diff;
233

234
	__ib_umem_release(umem->context->device, umem, 1);
235 236

	mm = get_task_mm(current);
237 238
	if (!mm) {
		kfree(umem);
239
		return;
240
	}
241

242 243
	diff = PAGE_ALIGN(umem->length + umem->offset) >> PAGE_SHIFT;

244 245 246 247 248
	/*
	 * We may be called with the mm's mmap_sem already held.  This
	 * can happen when a userspace munmap() is the call that drops
	 * the last reference to our file and calls our release
	 * method.  If there are memory regions to destroy, we'll end
249 250
	 * up here and not be able to take the mmap_sem.  In that case
	 * we defer the vm_locked accounting to the system workqueue.
251
	 */
252 253 254 255 256 257 258 259 260
	if (context->closing) {
		if (!down_write_trylock(&mm->mmap_sem)) {
			INIT_WORK(&umem->work, ib_umem_account);
			umem->mm   = mm;
			umem->diff = diff;

			schedule_work(&umem->work);
			return;
		}
261 262 263 264 265 266
	} else
		down_write(&mm->mmap_sem);

	current->mm->locked_vm -= diff;
	up_write(&mm->mmap_sem);
	mmput(mm);
267
	kfree(umem);
268 269 270 271 272 273 274 275 276 277 278
}
EXPORT_SYMBOL(ib_umem_release);

int ib_umem_page_count(struct ib_umem *umem)
{
	struct ib_umem_chunk *chunk;
	int shift;
	int i;
	int n;

	shift = ilog2(umem->page_size);
279

280 281 282 283
	n = 0;
	list_for_each_entry(chunk, &umem->chunk_list, list)
		for (i = 0; i < chunk->nmap; ++i)
			n += sg_dma_len(&chunk->page_list[i]) >> shift;
284

285
	return n;
286
}
287
EXPORT_SYMBOL(ib_umem_page_count);