1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /* Iterator helpers.
3 *
4 * Copyright (C) 2022 Red Hat, Inc. All Rights Reserved.
5 * Written by David Howells (dhowells@redhat.com)
6 */
7
8 #include <linux/export.h>
9 #include <linux/slab.h>
10 #include <linux/mm.h>
11 #include <linux/uio.h>
12 #include <linux/scatterlist.h>
13 #include <linux/netfs.h>
14 #include "internal.h"
15
16 /**
17 * netfs_extract_user_iter - Extract the pages from a user iterator into a bvec
18 * @orig: The original iterator
19 * @orig_len: The amount of iterator to copy
20 * @new: The iterator to be set up
21 * @extraction_flags: Flags to qualify the request
22 *
23 * Extract the page fragments from the given amount of the source iterator and
24 * build up a second iterator that refers to all of those bits. This allows
25 * the original iterator to disposed of.
26 *
27 * @extraction_flags can have ITER_ALLOW_P2PDMA set to request peer-to-peer DMA be
28 * allowed on the pages extracted.
29 *
30 * On success, the number of elements in the bvec is returned, the original
31 * iterator will have been advanced by the amount extracted.
32 *
33 * The iov_iter_extract_mode() function should be used to query how cleanup
34 * should be performed.
35 */
netfs_extract_user_iter(struct iov_iter * orig,size_t orig_len,struct iov_iter * new,iov_iter_extraction_t extraction_flags)36 ssize_t netfs_extract_user_iter(struct iov_iter *orig, size_t orig_len,
37 struct iov_iter *new,
38 iov_iter_extraction_t extraction_flags)
39 {
40 struct bio_vec *bv = NULL;
41 struct page **pages;
42 unsigned int cur_npages;
43 unsigned int max_pages;
44 unsigned int npages = 0;
45 unsigned int i;
46 ssize_t ret;
47 size_t count = orig_len, offset, len;
48 size_t bv_size, pg_size;
49
50 if (WARN_ON_ONCE(!iter_is_ubuf(orig) && !iter_is_iovec(orig)))
51 return -EIO;
52
53 max_pages = iov_iter_npages(orig, INT_MAX);
54 bv_size = array_size(max_pages, sizeof(*bv));
55 bv = kvmalloc(bv_size, GFP_KERNEL);
56 if (!bv)
57 return -ENOMEM;
58
59 /* Put the page list at the end of the bvec list storage. bvec
60 * elements are larger than page pointers, so as long as we work
61 * 0->last, we should be fine.
62 */
63 pg_size = array_size(max_pages, sizeof(*pages));
64 pages = (void *)bv + bv_size - pg_size;
65
66 while (count && npages < max_pages) {
67 ret = iov_iter_extract_pages(orig, &pages, count,
68 max_pages - npages, extraction_flags,
69 &offset);
70 if (ret < 0) {
71 pr_err("Couldn't get user pages (rc=%zd)\n", ret);
72 break;
73 }
74
75 if (ret > count) {
76 pr_err("get_pages rc=%zd more than %zu\n", ret, count);
77 break;
78 }
79
80 count -= ret;
81 ret += offset;
82 cur_npages = DIV_ROUND_UP(ret, PAGE_SIZE);
83
84 if (npages + cur_npages > max_pages) {
85 pr_err("Out of bvec array capacity (%u vs %u)\n",
86 npages + cur_npages, max_pages);
87 break;
88 }
89
90 for (i = 0; i < cur_npages; i++) {
91 len = ret > PAGE_SIZE ? PAGE_SIZE : ret;
92 bvec_set_page(bv + npages + i, *pages++, len - offset, offset);
93 ret -= len;
94 offset = 0;
95 }
96
97 npages += cur_npages;
98 }
99
100 iov_iter_bvec(new, orig->data_source, bv, npages, orig_len - count);
101 return npages;
102 }
103 EXPORT_SYMBOL_GPL(netfs_extract_user_iter);
104
105 /*
106 * Select the span of a bvec iterator we're going to use. Limit it by both maximum
107 * size and maximum number of segments. Returns the size of the span in bytes.
108 */
netfs_limit_bvec(const struct iov_iter * iter,size_t start_offset,size_t max_size,size_t max_segs)109 static size_t netfs_limit_bvec(const struct iov_iter *iter, size_t start_offset,
110 size_t max_size, size_t max_segs)
111 {
112 const struct bio_vec *bvecs = iter->bvec;
113 unsigned int nbv = iter->nr_segs, ix = 0, nsegs = 0;
114 size_t len, span = 0, n = iter->count;
115 size_t skip = iter->iov_offset + start_offset;
116
117 if (WARN_ON(!iov_iter_is_bvec(iter)) ||
118 WARN_ON(start_offset > n) ||
119 n == 0)
120 return 0;
121
122 while (n && ix < nbv && skip) {
123 len = bvecs[ix].bv_len;
124 if (skip < len)
125 break;
126 skip -= len;
127 n -= len;
128 ix++;
129 }
130
131 while (n && ix < nbv) {
132 len = min3(n, bvecs[ix].bv_len - skip, max_size);
133 span += len;
134 nsegs++;
135 ix++;
136 if (span >= max_size || nsegs >= max_segs)
137 break;
138 skip = 0;
139 n -= len;
140 }
141
142 return min(span, max_size);
143 }
144
145 /*
146 * Select the span of an xarray iterator we're going to use. Limit it by both
147 * maximum size and maximum number of segments. It is assumed that segments
148 * can be larger than a page in size, provided they're physically contiguous.
149 * Returns the size of the span in bytes.
150 */
netfs_limit_xarray(const struct iov_iter * iter,size_t start_offset,size_t max_size,size_t max_segs)151 static size_t netfs_limit_xarray(const struct iov_iter *iter, size_t start_offset,
152 size_t max_size, size_t max_segs)
153 {
154 struct folio *folio;
155 unsigned int nsegs = 0;
156 loff_t pos = iter->xarray_start + iter->iov_offset;
157 pgoff_t index = pos / PAGE_SIZE;
158 size_t span = 0, n = iter->count;
159
160 XA_STATE(xas, iter->xarray, index);
161
162 if (WARN_ON(!iov_iter_is_xarray(iter)) ||
163 WARN_ON(start_offset > n) ||
164 n == 0)
165 return 0;
166 max_size = min(max_size, n - start_offset);
167
168 rcu_read_lock();
169 xas_for_each(&xas, folio, ULONG_MAX) {
170 size_t offset, flen, len;
171 if (xas_retry(&xas, folio))
172 continue;
173 if (WARN_ON(xa_is_value(folio)))
174 break;
175 if (WARN_ON(folio_test_hugetlb(folio)))
176 break;
177
178 flen = folio_size(folio);
179 offset = offset_in_folio(folio, pos);
180 len = min(max_size, flen - offset);
181 span += len;
182 nsegs++;
183 if (span >= max_size || nsegs >= max_segs)
184 break;
185 }
186
187 rcu_read_unlock();
188 return min(span, max_size);
189 }
190
191 /*
192 * Select the span of a folio queue iterator we're going to use. Limit it by
193 * both maximum size and maximum number of segments. Returns the size of the
194 * span in bytes.
195 */
netfs_limit_folioq(const struct iov_iter * iter,size_t start_offset,size_t max_size,size_t max_segs)196 static size_t netfs_limit_folioq(const struct iov_iter *iter, size_t start_offset,
197 size_t max_size, size_t max_segs)
198 {
199 const struct folio_queue *folioq = iter->folioq;
200 unsigned int nsegs = 0;
201 unsigned int slot = iter->folioq_slot;
202 size_t span = 0, n = iter->count;
203
204 if (WARN_ON(!iov_iter_is_folioq(iter)) ||
205 WARN_ON(start_offset > n) ||
206 n == 0)
207 return 0;
208 max_size = umin(max_size, n - start_offset);
209
210 if (slot >= folioq_nr_slots(folioq)) {
211 folioq = folioq->next;
212 slot = 0;
213 }
214
215 start_offset += iter->iov_offset;
216 do {
217 size_t flen = folioq_folio_size(folioq, slot);
218
219 if (start_offset < flen) {
220 span += flen - start_offset;
221 nsegs++;
222 start_offset = 0;
223 } else {
224 start_offset -= flen;
225 }
226 if (span >= max_size || nsegs >= max_segs)
227 break;
228
229 slot++;
230 if (slot >= folioq_nr_slots(folioq)) {
231 folioq = folioq->next;
232 slot = 0;
233 }
234 } while (folioq);
235
236 return umin(span, max_size);
237 }
238
netfs_limit_iter(const struct iov_iter * iter,size_t start_offset,size_t max_size,size_t max_segs)239 size_t netfs_limit_iter(const struct iov_iter *iter, size_t start_offset,
240 size_t max_size, size_t max_segs)
241 {
242 if (iov_iter_is_folioq(iter))
243 return netfs_limit_folioq(iter, start_offset, max_size, max_segs);
244 if (iov_iter_is_bvec(iter))
245 return netfs_limit_bvec(iter, start_offset, max_size, max_segs);
246 if (iov_iter_is_xarray(iter))
247 return netfs_limit_xarray(iter, start_offset, max_size, max_segs);
248 BUG();
249 }
250 EXPORT_SYMBOL(netfs_limit_iter);
251