Root/fs/ntfs/attrib.c

1/**
2 * attrib.c - NTFS attribute operations. Part of the Linux-NTFS project.
3 *
4 * Copyright (c) 2001-2007 Anton Altaparmakov
5 * Copyright (c) 2002 Richard Russon
6 *
7 * This program/include file is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU General Public License as published
9 * by the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
11 *
12 * This program/include file is distributed in the hope that it will be
13 * useful, but WITHOUT ANY WARRANTY; without even the implied warranty
14 * of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this program (in the main directory of the Linux-NTFS
19 * distribution in the file COPYING); if not, write to the Free Software
20 * Foundation,Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
21 */
22
23#include <linux/buffer_head.h>
24#include <linux/sched.h>
25#include <linux/slab.h>
26#include <linux/swap.h>
27#include <linux/writeback.h>
28
29#include "attrib.h"
30#include "debug.h"
31#include "layout.h"
32#include "lcnalloc.h"
33#include "malloc.h"
34#include "mft.h"
35#include "ntfs.h"
36#include "types.h"
37
38/**
39 * ntfs_map_runlist_nolock - map (a part of) a runlist of an ntfs inode
40 * @ni: ntfs inode for which to map (part of) a runlist
41 * @vcn: map runlist part containing this vcn
42 * @ctx: active attribute search context if present or NULL if not
43 *
44 * Map the part of a runlist containing the @vcn of the ntfs inode @ni.
45 *
46 * If @ctx is specified, it is an active search context of @ni and its base mft
47 * record. This is needed when ntfs_map_runlist_nolock() encounters unmapped
48 * runlist fragments and allows their mapping. If you do not have the mft
49 * record mapped, you can specify @ctx as NULL and ntfs_map_runlist_nolock()
50 * will perform the necessary mapping and unmapping.
51 *
52 * Note, ntfs_map_runlist_nolock() saves the state of @ctx on entry and
53 * restores it before returning. Thus, @ctx will be left pointing to the same
54 * attribute on return as on entry. However, the actual pointers in @ctx may
55 * point to different memory locations on return, so you must remember to reset
56 * any cached pointers from the @ctx, i.e. after the call to
57 * ntfs_map_runlist_nolock(), you will probably want to do:
58 * m = ctx->mrec;
59 * a = ctx->attr;
60 * Assuming you cache ctx->attr in a variable @a of type ATTR_RECORD * and that
61 * you cache ctx->mrec in a variable @m of type MFT_RECORD *.
62 *
63 * Return 0 on success and -errno on error. There is one special error code
64 * which is not an error as such. This is -ENOENT. It means that @vcn is out
65 * of bounds of the runlist.
66 *
67 * Note the runlist can be NULL after this function returns if @vcn is zero and
68 * the attribute has zero allocated size, i.e. there simply is no runlist.
69 *
70 * WARNING: If @ctx is supplied, regardless of whether success or failure is
71 * returned, you need to check IS_ERR(@ctx->mrec) and if 'true' the @ctx
72 * is no longer valid, i.e. you need to either call
73 * ntfs_attr_reinit_search_ctx() or ntfs_attr_put_search_ctx() on it.
74 * In that case PTR_ERR(@ctx->mrec) will give you the error code for
75 * why the mapping of the old inode failed.
76 *
77 * Locking: - The runlist described by @ni must be locked for writing on entry
78 * and is locked on return. Note the runlist will be modified.
79 * - If @ctx is NULL, the base mft record of @ni must not be mapped on
80 * entry and it will be left unmapped on return.
81 * - If @ctx is not NULL, the base mft record must be mapped on entry
82 * and it will be left mapped on return.
83 */
84int ntfs_map_runlist_nolock(ntfs_inode *ni, VCN vcn, ntfs_attr_search_ctx *ctx)
85{
86    VCN end_vcn;
87    unsigned long flags;
88    ntfs_inode *base_ni;
89    MFT_RECORD *m;
90    ATTR_RECORD *a;
91    runlist_element *rl;
92    struct page *put_this_page = NULL;
93    int err = 0;
94    bool ctx_is_temporary, ctx_needs_reset;
95    ntfs_attr_search_ctx old_ctx = { NULL, };
96
97    ntfs_debug("Mapping runlist part containing vcn 0x%llx.",
98            (unsigned long long)vcn);
99    if (!NInoAttr(ni))
100        base_ni = ni;
101    else
102        base_ni = ni->ext.base_ntfs_ino;
103    if (!ctx) {
104        ctx_is_temporary = ctx_needs_reset = true;
105        m = map_mft_record(base_ni);
106        if (IS_ERR(m))
107            return PTR_ERR(m);
108        ctx = ntfs_attr_get_search_ctx(base_ni, m);
109        if (unlikely(!ctx)) {
110            err = -ENOMEM;
111            goto err_out;
112        }
113    } else {
114        VCN allocated_size_vcn;
115
116        BUG_ON(IS_ERR(ctx->mrec));
117        a = ctx->attr;
118        BUG_ON(!a->non_resident);
119        ctx_is_temporary = false;
120        end_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn);
121        read_lock_irqsave(&ni->size_lock, flags);
122        allocated_size_vcn = ni->allocated_size >>
123                ni->vol->cluster_size_bits;
124        read_unlock_irqrestore(&ni->size_lock, flags);
125        if (!a->data.non_resident.lowest_vcn && end_vcn <= 0)
126            end_vcn = allocated_size_vcn - 1;
127        /*
128         * If we already have the attribute extent containing @vcn in
129         * @ctx, no need to look it up again. We slightly cheat in
130         * that if vcn exceeds the allocated size, we will refuse to
131         * map the runlist below, so there is definitely no need to get
132         * the right attribute extent.
133         */
134        if (vcn >= allocated_size_vcn || (a->type == ni->type &&
135                a->name_length == ni->name_len &&
136                !memcmp((u8*)a + le16_to_cpu(a->name_offset),
137                ni->name, ni->name_len) &&
138                sle64_to_cpu(a->data.non_resident.lowest_vcn)
139                <= vcn && end_vcn >= vcn))
140            ctx_needs_reset = false;
141        else {
142            /* Save the old search context. */
143            old_ctx = *ctx;
144            /*
145             * If the currently mapped (extent) inode is not the
146             * base inode we will unmap it when we reinitialize the
147             * search context which means we need to get a
148             * reference to the page containing the mapped mft
149             * record so we do not accidentally drop changes to the
150             * mft record when it has not been marked dirty yet.
151             */
152            if (old_ctx.base_ntfs_ino && old_ctx.ntfs_ino !=
153                    old_ctx.base_ntfs_ino) {
154                put_this_page = old_ctx.ntfs_ino->page;
155                page_cache_get(put_this_page);
156            }
157            /*
158             * Reinitialize the search context so we can lookup the
159             * needed attribute extent.
160             */
161            ntfs_attr_reinit_search_ctx(ctx);
162            ctx_needs_reset = true;
163        }
164    }
165    if (ctx_needs_reset) {
166        err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
167                CASE_SENSITIVE, vcn, NULL, 0, ctx);
168        if (unlikely(err)) {
169            if (err == -ENOENT)
170                err = -EIO;
171            goto err_out;
172        }
173        BUG_ON(!ctx->attr->non_resident);
174    }
175    a = ctx->attr;
176    /*
177     * Only decompress the mapping pairs if @vcn is inside it. Otherwise
178     * we get into problems when we try to map an out of bounds vcn because
179     * we then try to map the already mapped runlist fragment and
180     * ntfs_mapping_pairs_decompress() fails.
181     */
182    end_vcn = sle64_to_cpu(a->data.non_resident.highest_vcn) + 1;
183    if (unlikely(vcn && vcn >= end_vcn)) {
184        err = -ENOENT;
185        goto err_out;
186    }
187    rl = ntfs_mapping_pairs_decompress(ni->vol, a, ni->runlist.rl);
188    if (IS_ERR(rl))
189        err = PTR_ERR(rl);
190    else
191        ni->runlist.rl = rl;
192err_out:
193    if (ctx_is_temporary) {
194        if (likely(ctx))
195            ntfs_attr_put_search_ctx(ctx);
196        unmap_mft_record(base_ni);
197    } else if (ctx_needs_reset) {
198        /*
199         * If there is no attribute list, restoring the search context
200         * is accomplished simply by copying the saved context back over
201         * the caller supplied context. If there is an attribute list,
202         * things are more complicated as we need to deal with mapping
203         * of mft records and resulting potential changes in pointers.
204         */
205        if (NInoAttrList(base_ni)) {
206            /*
207             * If the currently mapped (extent) inode is not the
208             * one we had before, we need to unmap it and map the
209             * old one.
210             */
211            if (ctx->ntfs_ino != old_ctx.ntfs_ino) {
212                /*
213                 * If the currently mapped inode is not the
214                 * base inode, unmap it.
215                 */
216                if (ctx->base_ntfs_ino && ctx->ntfs_ino !=
217                        ctx->base_ntfs_ino) {
218                    unmap_extent_mft_record(ctx->ntfs_ino);
219                    ctx->mrec = ctx->base_mrec;
220                    BUG_ON(!ctx->mrec);
221                }
222                /*
223                 * If the old mapped inode is not the base
224                 * inode, map it.
225                 */
226                if (old_ctx.base_ntfs_ino &&
227                        old_ctx.ntfs_ino !=
228                        old_ctx.base_ntfs_ino) {
229retry_map:
230                    ctx->mrec = map_mft_record(
231                            old_ctx.ntfs_ino);
232                    /*
233                     * Something bad has happened. If out
234                     * of memory retry till it succeeds.
235                     * Any other errors are fatal and we
236                     * return the error code in ctx->mrec.
237                     * Let the caller deal with it... We
238                     * just need to fudge things so the
239                     * caller can reinit and/or put the
240                     * search context safely.
241                     */
242                    if (IS_ERR(ctx->mrec)) {
243                        if (PTR_ERR(ctx->mrec) ==
244                                -ENOMEM) {
245                            schedule();
246                            goto retry_map;
247                        } else
248                            old_ctx.ntfs_ino =
249                                old_ctx.
250                                base_ntfs_ino;
251                    }
252                }
253            }
254            /* Update the changed pointers in the saved context. */
255            if (ctx->mrec != old_ctx.mrec) {
256                if (!IS_ERR(ctx->mrec))
257                    old_ctx.attr = (ATTR_RECORD*)(
258                            (u8*)ctx->mrec +
259                            ((u8*)old_ctx.attr -
260                            (u8*)old_ctx.mrec));
261                old_ctx.mrec = ctx->mrec;
262            }
263        }
264        /* Restore the search context to the saved one. */
265        *ctx = old_ctx;
266        /*
267         * We drop the reference on the page we took earlier. In the
268         * case that IS_ERR(ctx->mrec) is true this means we might lose
269         * some changes to the mft record that had been made between
270         * the last time it was marked dirty/written out and now. This
271         * at this stage is not a problem as the mapping error is fatal
272         * enough that the mft record cannot be written out anyway and
273         * the caller is very likely to shutdown the whole inode
274         * immediately and mark the volume dirty for chkdsk to pick up
275         * the pieces anyway.
276         */
277        if (put_this_page)
278            page_cache_release(put_this_page);
279    }
280    return err;
281}
282
283/**
284 * ntfs_map_runlist - map (a part of) a runlist of an ntfs inode
285 * @ni: ntfs inode for which to map (part of) a runlist
286 * @vcn: map runlist part containing this vcn
287 *
288 * Map the part of a runlist containing the @vcn of the ntfs inode @ni.
289 *
290 * Return 0 on success and -errno on error. There is one special error code
291 * which is not an error as such. This is -ENOENT. It means that @vcn is out
292 * of bounds of the runlist.
293 *
294 * Locking: - The runlist must be unlocked on entry and is unlocked on return.
295 * - This function takes the runlist lock for writing and may modify
296 * the runlist.
297 */
298int ntfs_map_runlist(ntfs_inode *ni, VCN vcn)
299{
300    int err = 0;
301
302    down_write(&ni->runlist.lock);
303    /* Make sure someone else didn't do the work while we were sleeping. */
304    if (likely(ntfs_rl_vcn_to_lcn(ni->runlist.rl, vcn) <=
305            LCN_RL_NOT_MAPPED))
306        err = ntfs_map_runlist_nolock(ni, vcn, NULL);
307    up_write(&ni->runlist.lock);
308    return err;
309}
310
311/**
312 * ntfs_attr_vcn_to_lcn_nolock - convert a vcn into a lcn given an ntfs inode
313 * @ni: ntfs inode of the attribute whose runlist to search
314 * @vcn: vcn to convert
315 * @write_locked: true if the runlist is locked for writing
316 *
317 * Find the virtual cluster number @vcn in the runlist of the ntfs attribute
318 * described by the ntfs inode @ni and return the corresponding logical cluster
319 * number (lcn).
320 *
321 * If the @vcn is not mapped yet, the attempt is made to map the attribute
322 * extent containing the @vcn and the vcn to lcn conversion is retried.
323 *
324 * If @write_locked is true the caller has locked the runlist for writing and
325 * if false for reading.
326 *
327 * Since lcns must be >= 0, we use negative return codes with special meaning:
328 *
329 * Return code Meaning / Description
330 * ==========================================
331 * LCN_HOLE Hole / not allocated on disk.
332 * LCN_ENOENT There is no such vcn in the runlist, i.e. @vcn is out of bounds.
333 * LCN_ENOMEM Not enough memory to map runlist.
334 * LCN_EIO Critical error (runlist/file is corrupt, i/o error, etc).
335 *
336 * Locking: - The runlist must be locked on entry and is left locked on return.
337 * - If @write_locked is 'false', i.e. the runlist is locked for reading,
338 * the lock may be dropped inside the function so you cannot rely on
339 * the runlist still being the same when this function returns.
340 */
341LCN ntfs_attr_vcn_to_lcn_nolock(ntfs_inode *ni, const VCN vcn,
342        const bool write_locked)
343{
344    LCN lcn;
345    unsigned long flags;
346    bool is_retry = false;
347
348    ntfs_debug("Entering for i_ino 0x%lx, vcn 0x%llx, %s_locked.",
349            ni->mft_no, (unsigned long long)vcn,
350            write_locked ? "write" : "read");
351    BUG_ON(!ni);
352    BUG_ON(!NInoNonResident(ni));
353    BUG_ON(vcn < 0);
354    if (!ni->runlist.rl) {
355        read_lock_irqsave(&ni->size_lock, flags);
356        if (!ni->allocated_size) {
357            read_unlock_irqrestore(&ni->size_lock, flags);
358            return LCN_ENOENT;
359        }
360        read_unlock_irqrestore(&ni->size_lock, flags);
361    }
362retry_remap:
363    /* Convert vcn to lcn. If that fails map the runlist and retry once. */
364    lcn = ntfs_rl_vcn_to_lcn(ni->runlist.rl, vcn);
365    if (likely(lcn >= LCN_HOLE)) {
366        ntfs_debug("Done, lcn 0x%llx.", (long long)lcn);
367        return lcn;
368    }
369    if (lcn != LCN_RL_NOT_MAPPED) {
370        if (lcn != LCN_ENOENT)
371            lcn = LCN_EIO;
372    } else if (!is_retry) {
373        int err;
374
375        if (!write_locked) {
376            up_read(&ni->runlist.lock);
377            down_write(&ni->runlist.lock);
378            if (unlikely(ntfs_rl_vcn_to_lcn(ni->runlist.rl, vcn) !=
379                    LCN_RL_NOT_MAPPED)) {
380                up_write(&ni->runlist.lock);
381                down_read(&ni->runlist.lock);
382                goto retry_remap;
383            }
384        }
385        err = ntfs_map_runlist_nolock(ni, vcn, NULL);
386        if (!write_locked) {
387            up_write(&ni->runlist.lock);
388            down_read(&ni->runlist.lock);
389        }
390        if (likely(!err)) {
391            is_retry = true;
392            goto retry_remap;
393        }
394        if (err == -ENOENT)
395            lcn = LCN_ENOENT;
396        else if (err == -ENOMEM)
397            lcn = LCN_ENOMEM;
398        else
399            lcn = LCN_EIO;
400    }
401    if (lcn != LCN_ENOENT)
402        ntfs_error(ni->vol->sb, "Failed with error code %lli.",
403                (long long)lcn);
404    return lcn;
405}
406
407/**
408 * ntfs_attr_find_vcn_nolock - find a vcn in the runlist of an ntfs inode
409 * @ni: ntfs inode describing the runlist to search
410 * @vcn: vcn to find
411 * @ctx: active attribute search context if present or NULL if not
412 *
413 * Find the virtual cluster number @vcn in the runlist described by the ntfs
414 * inode @ni and return the address of the runlist element containing the @vcn.
415 *
416 * If the @vcn is not mapped yet, the attempt is made to map the attribute
417 * extent containing the @vcn and the vcn to lcn conversion is retried.
418 *
419 * If @ctx is specified, it is an active search context of @ni and its base mft
420 * record. This is needed when ntfs_attr_find_vcn_nolock() encounters unmapped
421 * runlist fragments and allows their mapping. If you do not have the mft
422 * record mapped, you can specify @ctx as NULL and ntfs_attr_find_vcn_nolock()
423 * will perform the necessary mapping and unmapping.
424 *
425 * Note, ntfs_attr_find_vcn_nolock() saves the state of @ctx on entry and
426 * restores it before returning. Thus, @ctx will be left pointing to the same
427 * attribute on return as on entry. However, the actual pointers in @ctx may
428 * point to different memory locations on return, so you must remember to reset
429 * any cached pointers from the @ctx, i.e. after the call to
430 * ntfs_attr_find_vcn_nolock(), you will probably want to do:
431 * m = ctx->mrec;
432 * a = ctx->attr;
433 * Assuming you cache ctx->attr in a variable @a of type ATTR_RECORD * and that
434 * you cache ctx->mrec in a variable @m of type MFT_RECORD *.
435 * Note you need to distinguish between the lcn of the returned runlist element
436 * being >= 0 and LCN_HOLE. In the later case you have to return zeroes on
437 * read and allocate clusters on write.
438 *
439 * Return the runlist element containing the @vcn on success and
440 * ERR_PTR(-errno) on error. You need to test the return value with IS_ERR()
441 * to decide if the return is success or failure and PTR_ERR() to get to the
442 * error code if IS_ERR() is true.
443 *
444 * The possible error return codes are:
445 * -ENOENT - No such vcn in the runlist, i.e. @vcn is out of bounds.
446 * -ENOMEM - Not enough memory to map runlist.
447 * -EIO - Critical error (runlist/file is corrupt, i/o error, etc).
448 *
449 * WARNING: If @ctx is supplied, regardless of whether success or failure is
450 * returned, you need to check IS_ERR(@ctx->mrec) and if 'true' the @ctx
451 * is no longer valid, i.e. you need to either call
452 * ntfs_attr_reinit_search_ctx() or ntfs_attr_put_search_ctx() on it.
453 * In that case PTR_ERR(@ctx->mrec) will give you the error code for
454 * why the mapping of the old inode failed.
455 *
456 * Locking: - The runlist described by @ni must be locked for writing on entry
457 * and is locked on return. Note the runlist may be modified when
458 * needed runlist fragments need to be mapped.
459 * - If @ctx is NULL, the base mft record of @ni must not be mapped on
460 * entry and it will be left unmapped on return.
461 * - If @ctx is not NULL, the base mft record must be mapped on entry
462 * and it will be left mapped on return.
463 */
464runlist_element *ntfs_attr_find_vcn_nolock(ntfs_inode *ni, const VCN vcn,
465        ntfs_attr_search_ctx *ctx)
466{
467    unsigned long flags;
468    runlist_element *rl;
469    int err = 0;
470    bool is_retry = false;
471
472    ntfs_debug("Entering for i_ino 0x%lx, vcn 0x%llx, with%s ctx.",
473            ni->mft_no, (unsigned long long)vcn, ctx ? "" : "out");
474    BUG_ON(!ni);
475    BUG_ON(!NInoNonResident(ni));
476    BUG_ON(vcn < 0);
477    if (!ni->runlist.rl) {
478        read_lock_irqsave(&ni->size_lock, flags);
479        if (!ni->allocated_size) {
480            read_unlock_irqrestore(&ni->size_lock, flags);
481            return ERR_PTR(-ENOENT);
482        }
483        read_unlock_irqrestore(&ni->size_lock, flags);
484    }
485retry_remap:
486    rl = ni->runlist.rl;
487    if (likely(rl && vcn >= rl[0].vcn)) {
488        while (likely(rl->length)) {
489            if (unlikely(vcn < rl[1].vcn)) {
490                if (likely(rl->lcn >= LCN_HOLE)) {
491                    ntfs_debug("Done.");
492                    return rl;
493                }
494                break;
495            }
496            rl++;
497        }
498        if (likely(rl->lcn != LCN_RL_NOT_MAPPED)) {
499            if (likely(rl->lcn == LCN_ENOENT))
500                err = -ENOENT;
501            else
502                err = -EIO;
503        }
504    }
505    if (!err && !is_retry) {
506        /*
507         * If the search context is invalid we cannot map the unmapped
508         * region.
509         */
510        if (IS_ERR(ctx->mrec))
511            err = PTR_ERR(ctx->mrec);
512        else {
513            /*
514             * The @vcn is in an unmapped region, map the runlist
515             * and retry.
516             */
517            err = ntfs_map_runlist_nolock(ni, vcn, ctx);
518            if (likely(!err)) {
519                is_retry = true;
520                goto retry_remap;
521            }
522        }
523        if (err == -EINVAL)
524            err = -EIO;
525    } else if (!err)
526        err = -EIO;
527    if (err != -ENOENT)
528        ntfs_error(ni->vol->sb, "Failed with error code %i.", err);
529    return ERR_PTR(err);
530}
531
532/**
533 * ntfs_attr_find - find (next) attribute in mft record
534 * @type: attribute type to find
535 * @name: attribute name to find (optional, i.e. NULL means don't care)
536 * @name_len: attribute name length (only needed if @name present)
537 * @ic: IGNORE_CASE or CASE_SENSITIVE (ignored if @name not present)
538 * @val: attribute value to find (optional, resident attributes only)
539 * @val_len: attribute value length
540 * @ctx: search context with mft record and attribute to search from
541 *
542 * You should not need to call this function directly. Use ntfs_attr_lookup()
543 * instead.
544 *
545 * ntfs_attr_find() takes a search context @ctx as parameter and searches the
546 * mft record specified by @ctx->mrec, beginning at @ctx->attr, for an
547 * attribute of @type, optionally @name and @val.
548 *
549 * If the attribute is found, ntfs_attr_find() returns 0 and @ctx->attr will
550 * point to the found attribute.
551 *
552 * If the attribute is not found, ntfs_attr_find() returns -ENOENT and
553 * @ctx->attr will point to the attribute before which the attribute being
554 * searched for would need to be inserted if such an action were to be desired.
555 *
556 * On actual error, ntfs_attr_find() returns -EIO. In this case @ctx->attr is
557 * undefined and in particular do not rely on it not changing.
558 *
559 * If @ctx->is_first is 'true', the search begins with @ctx->attr itself. If it
560 * is 'false', the search begins after @ctx->attr.
561 *
562 * If @ic is IGNORE_CASE, the @name comparisson is not case sensitive and
563 * @ctx->ntfs_ino must be set to the ntfs inode to which the mft record
564 * @ctx->mrec belongs. This is so we can get at the ntfs volume and hence at
565 * the upcase table. If @ic is CASE_SENSITIVE, the comparison is case
566 * sensitive. When @name is present, @name_len is the @name length in Unicode
567 * characters.
568 *
569 * If @name is not present (NULL), we assume that the unnamed attribute is
570 * being searched for.
571 *
572 * Finally, the resident attribute value @val is looked for, if present. If
573 * @val is not present (NULL), @val_len is ignored.
574 *
575 * ntfs_attr_find() only searches the specified mft record and it ignores the
576 * presence of an attribute list attribute (unless it is the one being searched
577 * for, obviously). If you need to take attribute lists into consideration,
578 * use ntfs_attr_lookup() instead (see below). This also means that you cannot
579 * use ntfs_attr_find() to search for extent records of non-resident
580 * attributes, as extents with lowest_vcn != 0 are usually described by the
581 * attribute list attribute only. - Note that it is possible that the first
582 * extent is only in the attribute list while the last extent is in the base
583 * mft record, so do not rely on being able to find the first extent in the
584 * base mft record.
585 *
586 * Warning: Never use @val when looking for attribute types which can be
587 * non-resident as this most likely will result in a crash!
588 */
589static int ntfs_attr_find(const ATTR_TYPE type, const ntfschar *name,
590        const u32 name_len, const IGNORE_CASE_BOOL ic,
591        const u8 *val, const u32 val_len, ntfs_attr_search_ctx *ctx)
592{
593    ATTR_RECORD *a;
594    ntfs_volume *vol = ctx->ntfs_ino->vol;
595    ntfschar *upcase = vol->upcase;
596    u32 upcase_len = vol->upcase_len;
597
598    /*
599     * Iterate over attributes in mft record starting at @ctx->attr, or the
600     * attribute following that, if @ctx->is_first is 'true'.
601     */
602    if (ctx->is_first) {
603        a = ctx->attr;
604        ctx->is_first = false;
605    } else
606        a = (ATTR_RECORD*)((u8*)ctx->attr +
607                le32_to_cpu(ctx->attr->length));
608    for (;; a = (ATTR_RECORD*)((u8*)a + le32_to_cpu(a->length))) {
609        if ((u8*)a < (u8*)ctx->mrec || (u8*)a > (u8*)ctx->mrec +
610                le32_to_cpu(ctx->mrec->bytes_allocated))
611            break;
612        ctx->attr = a;
613        if (unlikely(le32_to_cpu(a->type) > le32_to_cpu(type) ||
614                a->type == AT_END))
615            return -ENOENT;
616        if (unlikely(!a->length))
617            break;
618        if (a->type != type)
619            continue;
620        /*
621         * If @name is present, compare the two names. If @name is
622         * missing, assume we want an unnamed attribute.
623         */
624        if (!name) {
625            /* The search failed if the found attribute is named. */
626            if (a->name_length)
627                return -ENOENT;
628        } else if (!ntfs_are_names_equal(name, name_len,
629                (ntfschar*)((u8*)a + le16_to_cpu(a->name_offset)),
630                a->name_length, ic, upcase, upcase_len)) {
631            register int rc;
632
633            rc = ntfs_collate_names(name, name_len,
634                    (ntfschar*)((u8*)a +
635                    le16_to_cpu(a->name_offset)),
636                    a->name_length, 1, IGNORE_CASE,
637                    upcase, upcase_len);
638            /*
639             * If @name collates before a->name, there is no
640             * matching attribute.
641             */
642            if (rc == -1)
643                return -ENOENT;
644            /* If the strings are not equal, continue search. */
645            if (rc)
646                continue;
647            rc = ntfs_collate_names(name, name_len,
648                    (ntfschar*)((u8*)a +
649                    le16_to_cpu(a->name_offset)),
650                    a->name_length, 1, CASE_SENSITIVE,
651                    upcase, upcase_len);
652            if (rc == -1)
653                return -ENOENT;
654            if (rc)
655                continue;
656        }
657        /*
658         * The names match or @name not present and attribute is
659         * unnamed. If no @val specified, we have found the attribute
660         * and are done.
661         */
662        if (!val)
663            return 0;
664        /* @val is present; compare values. */
665        else {
666            register int rc;
667
668            rc = memcmp(val, (u8*)a + le16_to_cpu(
669                    a->data.resident.value_offset),
670                    min_t(u32, val_len, le32_to_cpu(
671                    a->data.resident.value_length)));
672            /*
673             * If @val collates before the current attribute's
674             * value, there is no matching attribute.
675             */
676            if (!rc) {
677                register u32 avl;
678
679                avl = le32_to_cpu(
680                        a->data.resident.value_length);
681                if (val_len == avl)
682                    return 0;
683                if (val_len < avl)
684                    return -ENOENT;
685            } else if (rc < 0)
686                return -ENOENT;
687        }
688    }
689    ntfs_error(vol->sb, "Inode is corrupt. Run chkdsk.");
690    NVolSetErrors(vol);
691    return -EIO;
692}
693
694/**
695 * load_attribute_list - load an attribute list into memory
696 * @vol: ntfs volume from which to read
697 * @runlist: runlist of the attribute list
698 * @al_start: destination buffer
699 * @size: size of the destination buffer in bytes
700 * @initialized_size: initialized size of the attribute list
701 *
702 * Walk the runlist @runlist and load all clusters from it copying them into
703 * the linear buffer @al. The maximum number of bytes copied to @al is @size
704 * bytes. Note, @size does not need to be a multiple of the cluster size. If
705 * @initialized_size is less than @size, the region in @al between
706 * @initialized_size and @size will be zeroed and not read from disk.
707 *
708 * Return 0 on success or -errno on error.
709 */
710int load_attribute_list(ntfs_volume *vol, runlist *runlist, u8 *al_start,
711        const s64 size, const s64 initialized_size)
712{
713    LCN lcn;
714    u8 *al = al_start;
715    u8 *al_end = al + initialized_size;
716    runlist_element *rl;
717    struct buffer_head *bh;
718    struct super_block *sb;
719    unsigned long block_size;
720    unsigned long block, max_block;
721    int err = 0;
722    unsigned char block_size_bits;
723
724    ntfs_debug("Entering.");
725    if (!vol || !runlist || !al || size <= 0 || initialized_size < 0 ||
726            initialized_size > size)
727        return -EINVAL;
728    if (!initialized_size) {
729        memset(al, 0, size);
730        return 0;
731    }
732    sb = vol->sb;
733    block_size = sb->s_blocksize;
734    block_size_bits = sb->s_blocksize_bits;
735    down_read(&runlist->lock);
736    rl = runlist->rl;
737    if (!rl) {
738        ntfs_error(sb, "Cannot read attribute list since runlist is "
739                "missing.");
740        goto err_out;
741    }
742    /* Read all clusters specified by the runlist one run at a time. */
743    while (rl->length) {
744        lcn = ntfs_rl_vcn_to_lcn(rl, rl->vcn);
745        ntfs_debug("Reading vcn = 0x%llx, lcn = 0x%llx.",
746                (unsigned long long)rl->vcn,
747                (unsigned long long)lcn);
748        /* The attribute list cannot be sparse. */
749        if (lcn < 0) {
750            ntfs_error(sb, "ntfs_rl_vcn_to_lcn() failed. Cannot "
751                    "read attribute list.");
752            goto err_out;
753        }
754        block = lcn << vol->cluster_size_bits >> block_size_bits;
755        /* Read the run from device in chunks of block_size bytes. */
756        max_block = block + (rl->length << vol->cluster_size_bits >>
757                block_size_bits);
758        ntfs_debug("max_block = 0x%lx.", max_block);
759        do {
760            ntfs_debug("Reading block = 0x%lx.", block);
761            bh = sb_bread(sb, block);
762            if (!bh) {
763                ntfs_error(sb, "sb_bread() failed. Cannot "
764                        "read attribute list.");
765                goto err_out;
766            }
767            if (al + block_size >= al_end)
768                goto do_final;
769            memcpy(al, bh->b_data, block_size);
770            brelse(bh);
771            al += block_size;
772        } while (++block < max_block);
773        rl++;
774    }
775    if (initialized_size < size) {
776initialize:
777        memset(al_start + initialized_size, 0, size - initialized_size);
778    }
779done:
780    up_read(&runlist->lock);
781    return err;
782do_final:
783    if (al < al_end) {
784        /*
785         * Partial block.
786         *
787         * Note: The attribute list can be smaller than its allocation
788         * by multiple clusters. This has been encountered by at least
789         * two people running Windows XP, thus we cannot do any
790         * truncation sanity checking here. (AIA)
791         */
792        memcpy(al, bh->b_data, al_end - al);
793        brelse(bh);
794        if (initialized_size < size)
795            goto initialize;
796        goto done;
797    }
798    brelse(bh);
799    /* Real overflow! */
800    ntfs_error(sb, "Attribute list buffer overflow. Read attribute list "
801            "is truncated.");
802err_out:
803    err = -EIO;
804    goto done;
805}
806
807/**
808 * ntfs_external_attr_find - find an attribute in the attribute list of an inode
809 * @type: attribute type to find
810 * @name: attribute name to find (optional, i.e. NULL means don't care)
811 * @name_len: attribute name length (only needed if @name present)
812 * @ic: IGNORE_CASE or CASE_SENSITIVE (ignored if @name not present)
813 * @lowest_vcn: lowest vcn to find (optional, non-resident attributes only)
814 * @val: attribute value to find (optional, resident attributes only)
815 * @val_len: attribute value length
816 * @ctx: search context with mft record and attribute to search from
817 *
818 * You should not need to call this function directly. Use ntfs_attr_lookup()
819 * instead.
820 *
821 * Find an attribute by searching the attribute list for the corresponding
822 * attribute list entry. Having found the entry, map the mft record if the
823 * attribute is in a different mft record/inode, ntfs_attr_find() the attribute
824 * in there and return it.
825 *
826 * On first search @ctx->ntfs_ino must be the base mft record and @ctx must
827 * have been obtained from a call to ntfs_attr_get_search_ctx(). On subsequent
828 * calls @ctx->ntfs_ino can be any extent inode, too (@ctx->base_ntfs_ino is
829 * then the base inode).
830 *
831 * After finishing with the attribute/mft record you need to call
832 * ntfs_attr_put_search_ctx() to cleanup the search context (unmapping any
833 * mapped inodes, etc).
834 *
835 * If the attribute is found, ntfs_external_attr_find() returns 0 and
836 * @ctx->attr will point to the found attribute. @ctx->mrec will point to the
837 * mft record in which @ctx->attr is located and @ctx->al_entry will point to
838 * the attribute list entry for the attribute.
839 *
840 * If the attribute is not found, ntfs_external_attr_find() returns -ENOENT and
841 * @ctx->attr will point to the attribute in the base mft record before which
842 * the attribute being searched for would need to be inserted if such an action
843 * were to be desired. @ctx->mrec will point to the mft record in which
844 * @ctx->attr is located and @ctx->al_entry will point to the attribute list
845 * entry of the attribute before which the attribute being searched for would
846 * need to be inserted if such an action were to be desired.
847 *
848 * Thus to insert the not found attribute, one wants to add the attribute to
849 * @ctx->mrec (the base mft record) and if there is not enough space, the
850 * attribute should be placed in a newly allocated extent mft record. The
851 * attribute list entry for the inserted attribute should be inserted in the
852 * attribute list attribute at @ctx->al_entry.
853 *
854 * On actual error, ntfs_external_attr_find() returns -EIO. In this case
855 * @ctx->attr is undefined and in particular do not rely on it not changing.
856 */
857static int ntfs_external_attr_find(const ATTR_TYPE type,
858        const ntfschar *name, const u32 name_len,
859        const IGNORE_CASE_BOOL ic, const VCN lowest_vcn,
860        const u8 *val, const u32 val_len, ntfs_attr_search_ctx *ctx)
861{
862    ntfs_inode *base_ni, *ni;
863    ntfs_volume *vol;
864    ATTR_LIST_ENTRY *al_entry, *next_al_entry;
865    u8 *al_start, *al_end;
866    ATTR_RECORD *a;
867    ntfschar *al_name;
868    u32 al_name_len;
869    int err = 0;
870    static const char *es = " Unmount and run chkdsk.";
871
872    ni = ctx->ntfs_ino;
873    base_ni = ctx->base_ntfs_ino;
874    ntfs_debug("Entering for inode 0x%lx, type 0x%x.", ni->mft_no, type);
875    if (!base_ni) {
876        /* First call happens with the base mft record. */
877        base_ni = ctx->base_ntfs_ino = ctx->ntfs_ino;
878        ctx->base_mrec = ctx->mrec;
879    }
880    if (ni == base_ni)
881        ctx->base_attr = ctx->attr;
882    if (type == AT_END)
883        goto not_found;
884    vol = base_ni->vol;
885    al_start = base_ni->attr_list;
886    al_end = al_start + base_ni->attr_list_size;
887    if (!ctx->al_entry)
888        ctx->al_entry = (ATTR_LIST_ENTRY*)al_start;
889    /*
890     * Iterate over entries in attribute list starting at @ctx->al_entry,
891     * or the entry following that, if @ctx->is_first is 'true'.
892     */
893    if (ctx->is_first) {
894        al_entry = ctx->al_entry;
895        ctx->is_first = false;
896    } else
897        al_entry = (ATTR_LIST_ENTRY*)((u8*)ctx->al_entry +
898                le16_to_cpu(ctx->al_entry->length));
899    for (;; al_entry = next_al_entry) {
900        /* Out of bounds check. */
901        if ((u8*)al_entry < base_ni->attr_list ||
902                (u8*)al_entry > al_end)
903            break; /* Inode is corrupt. */
904        ctx->al_entry = al_entry;
905        /* Catch the end of the attribute list. */
906        if ((u8*)al_entry == al_end)
907            goto not_found;
908        if (!al_entry->length)
909            break;
910        if ((u8*)al_entry + 6 > al_end || (u8*)al_entry +
911                le16_to_cpu(al_entry->length) > al_end)
912            break;
913        next_al_entry = (ATTR_LIST_ENTRY*)((u8*)al_entry +
914                le16_to_cpu(al_entry->length));
915        if (le32_to_cpu(al_entry->type) > le32_to_cpu(type))
916            goto not_found;
917        if (type != al_entry->type)
918            continue;
919        /*
920         * If @name is present, compare the two names. If @name is
921         * missing, assume we want an unnamed attribute.
922         */
923        al_name_len = al_entry->name_length;
924        al_name = (ntfschar*)((u8*)al_entry + al_entry->name_offset);
925        if (!name) {
926            if (al_name_len)
927                goto not_found;
928        } else if (!ntfs_are_names_equal(al_name, al_name_len, name,
929                name_len, ic, vol->upcase, vol->upcase_len)) {
930            register int rc;
931
932            rc = ntfs_collate_names(name, name_len, al_name,
933                    al_name_len, 1, IGNORE_CASE,
934                    vol->upcase, vol->upcase_len);
935            /*
936             * If @name collates before al_name, there is no
937             * matching attribute.
938             */
939            if (rc == -1)
940                goto not_found;
941            /* If the strings are not equal, continue search. */
942            if (rc)
943                continue;
944            /*
945             * FIXME: Reverse engineering showed 0, IGNORE_CASE but
946             * that is inconsistent with ntfs_attr_find(). The
947             * subsequent rc checks were also different. Perhaps I
948             * made a mistake in one of the two. Need to recheck
949             * which is correct or at least see what is going on...
950             * (AIA)
951             */
952            rc = ntfs_collate_names(name, name_len, al_name,
953                    al_name_len, 1, CASE_SENSITIVE,
954                    vol->upcase, vol->upcase_len);
955            if (rc == -1)
956                goto not_found;
957            if (rc)
958                continue;
959        }
960        /*
961         * The names match or @name not present and attribute is
962         * unnamed. Now check @lowest_vcn. Continue search if the
963         * next attribute list entry still fits @lowest_vcn. Otherwise
964         * we have reached the right one or the search has failed.
965         */
966        if (lowest_vcn && (u8*)next_al_entry >= al_start &&
967                (u8*)next_al_entry + 6 < al_end &&
968                (u8*)next_al_entry + le16_to_cpu(
969                    next_al_entry->length) <= al_end &&
970                sle64_to_cpu(next_al_entry->lowest_vcn) <=
971                    lowest_vcn &&
972                next_al_entry->type == al_entry->type &&
973                next_al_entry->name_length == al_name_len &&
974                ntfs_are_names_equal((ntfschar*)((u8*)
975                    next_al_entry +
976                    next_al_entry->name_offset),
977                    next_al_entry->name_length,
978                    al_name, al_name_len, CASE_SENSITIVE,
979                    vol->upcase, vol->upcase_len))
980            continue;
981        if (MREF_LE(al_entry->mft_reference) == ni->mft_no) {
982            if (MSEQNO_LE(al_entry->mft_reference) != ni->seq_no) {
983                ntfs_error(vol->sb, "Found stale mft "
984                        "reference in attribute list "
985                        "of base inode 0x%lx.%s",
986                        base_ni->mft_no, es);
987                err = -EIO;
988                break;
989            }
990        } else { /* Mft references do not match. */
991            /* If there is a mapped record unmap it first. */
992            if (ni != base_ni)
993                unmap_extent_mft_record(ni);
994            /* Do we want the base record back? */
995            if (MREF_LE(al_entry->mft_reference) ==
996                    base_ni->mft_no) {
997                ni = ctx->ntfs_ino = base_ni;
998                ctx->mrec = ctx->base_mrec;
999            } else {
1000                /* We want an extent record. */
1001                ctx->mrec = map_extent_mft_record(base_ni,
1002                        le64_to_cpu(
1003                        al_entry->mft_reference), &ni);
1004                if (IS_ERR(ctx->mrec)) {
1005                    ntfs_error(vol->sb, "Failed to map "
1006                            "extent mft record "
1007                            "0x%lx of base inode "
1008                            "0x%lx.%s",
1009                            MREF_LE(al_entry->
1010                            mft_reference),
1011                            base_ni->mft_no, es);
1012                    err = PTR_ERR(ctx->mrec);
1013                    if (err == -ENOENT)
1014                        err = -EIO;
1015                    /* Cause @ctx to be sanitized below. */
1016                    ni = NULL;
1017                    break;
1018                }
1019                ctx->ntfs_ino = ni;
1020            }
1021            ctx->attr = (ATTR_RECORD*)((u8*)ctx->mrec +
1022                    le16_to_cpu(ctx->mrec->attrs_offset));
1023        }
1024        /*
1025         * ctx->vfs_ino, ctx->mrec, and ctx->attr now point to the
1026         * mft record containing the attribute represented by the
1027         * current al_entry.
1028         */
1029        /*
1030         * We could call into ntfs_attr_find() to find the right
1031         * attribute in this mft record but this would be less
1032         * efficient and not quite accurate as ntfs_attr_find() ignores
1033         * the attribute instance numbers for example which become
1034         * important when one plays with attribute lists. Also,
1035         * because a proper match has been found in the attribute list
1036         * entry above, the comparison can now be optimized. So it is
1037         * worth re-implementing a simplified ntfs_attr_find() here.
1038         */
1039        a = ctx->attr;
1040        /*
1041         * Use a manual loop so we can still use break and continue
1042         * with the same meanings as above.
1043         */
1044do_next_attr_loop:
1045        if ((u8*)a < (u8*)ctx->mrec || (u8*)a > (u8*)ctx->mrec +
1046                le32_to_cpu(ctx->mrec->bytes_allocated))
1047            break;
1048        if (a->type == AT_END)
1049            break;
1050        if (!a->length)
1051            break;
1052        if (al_entry->instance != a->instance)
1053            goto do_next_attr;
1054        /*
1055         * If the type and/or the name are mismatched between the
1056         * attribute list entry and the attribute record, there is
1057         * corruption so we break and return error EIO.
1058         */
1059        if (al_entry->type != a->type)
1060            break;
1061        if (!ntfs_are_names_equal((ntfschar*)((u8*)a +
1062                le16_to_cpu(a->name_offset)), a->name_length,
1063                al_name, al_name_len, CASE_SENSITIVE,
1064                vol->upcase, vol->upcase_len))
1065            break;
1066        ctx->attr = a;
1067        /*
1068         * If no @val specified or @val specified and it matches, we
1069         * have found it!
1070         */
1071        if (!val || (!a->non_resident && le32_to_cpu(
1072                a->data.resident.value_length) == val_len &&
1073                !memcmp((u8*)a +
1074                le16_to_cpu(a->data.resident.value_offset),
1075                val, val_len))) {
1076            ntfs_debug("Done, found.");
1077            return 0;
1078        }
1079do_next_attr:
1080        /* Proceed to the next attribute in the current mft record. */
1081        a = (ATTR_RECORD*)((u8*)a + le32_to_cpu(a->length));
1082        goto do_next_attr_loop;
1083    }
1084    if (!err) {
1085        ntfs_error(vol->sb, "Base inode 0x%lx contains corrupt "
1086                "attribute list attribute.%s", base_ni->mft_no,
1087                es);
1088        err = -EIO;
1089    }
1090    if (ni != base_ni) {
1091        if (ni)
1092            unmap_extent_mft_record(ni);
1093        ctx->ntfs_ino = base_ni;
1094        ctx->mrec = ctx->base_mrec;
1095        ctx->attr = ctx->base_attr;
1096    }
1097    if (err != -ENOMEM)
1098        NVolSetErrors(vol);
1099    return err;
1100not_found:
1101    /*
1102     * If we were looking for AT_END, we reset the search context @ctx and
1103     * use ntfs_attr_find() to seek to the end of the base mft record.
1104     */
1105    if (type == AT_END) {
1106        ntfs_attr_reinit_search_ctx(ctx);
1107        return ntfs_attr_find(AT_END, name, name_len, ic, val, val_len,
1108                ctx);
1109    }
1110    /*
1111     * The attribute was not found. Before we return, we want to ensure
1112     * @ctx->mrec and @ctx->attr indicate the position at which the
1113     * attribute should be inserted in the base mft record. Since we also
1114     * want to preserve @ctx->al_entry we cannot reinitialize the search
1115     * context using ntfs_attr_reinit_search_ctx() as this would set
1116     * @ctx->al_entry to NULL. Thus we do the necessary bits manually (see
1117     * ntfs_attr_init_search_ctx() below). Note, we _only_ preserve
1118     * @ctx->al_entry as the remaining fields (base_*) are identical to
1119     * their non base_ counterparts and we cannot set @ctx->base_attr
1120     * correctly yet as we do not know what @ctx->attr will be set to by
1121     * the call to ntfs_attr_find() below.
1122     */
1123    if (ni != base_ni)
1124        unmap_extent_mft_record(ni);
1125    ctx->mrec = ctx->base_mrec;
1126    ctx->attr = (ATTR_RECORD*)((u8*)ctx->mrec +
1127            le16_to_cpu(ctx->mrec->attrs_offset));
1128    ctx->is_first = true;
1129    ctx->ntfs_ino = base_ni;
1130    ctx->base_ntfs_ino = NULL;
1131    ctx->base_mrec = NULL;
1132    ctx->base_attr = NULL;
1133    /*
1134     * In case there are multiple matches in the base mft record, need to
1135     * keep enumerating until we get an attribute not found response (or
1136     * another error), otherwise we would keep returning the same attribute
1137     * over and over again and all programs using us for enumeration would
1138     * lock up in a tight loop.
1139     */
1140    do {
1141        err = ntfs_attr_find(type, name, name_len, ic, val, val_len,
1142                ctx);
1143    } while (!err);
1144    ntfs_debug("Done, not found.");
1145    return err;
1146}
1147
1148/**
1149 * ntfs_attr_lookup - find an attribute in an ntfs inode
1150 * @type: attribute type to find
1151 * @name: attribute name to find (optional, i.e. NULL means don't care)
1152 * @name_len: attribute name length (only needed if @name present)
1153 * @ic: IGNORE_CASE or CASE_SENSITIVE (ignored if @name not present)
1154 * @lowest_vcn: lowest vcn to find (optional, non-resident attributes only)
1155 * @val: attribute value to find (optional, resident attributes only)
1156 * @val_len: attribute value length
1157 * @ctx: search context with mft record and attribute to search from
1158 *
1159 * Find an attribute in an ntfs inode. On first search @ctx->ntfs_ino must
1160 * be the base mft record and @ctx must have been obtained from a call to
1161 * ntfs_attr_get_search_ctx().
1162 *
1163 * This function transparently handles attribute lists and @ctx is used to
1164 * continue searches where they were left off at.
1165 *
1166 * After finishing with the attribute/mft record you need to call
1167 * ntfs_attr_put_search_ctx() to cleanup the search context (unmapping any
1168 * mapped inodes, etc).
1169 *
1170 * Return 0 if the search was successful and -errno if not.
1171 *
1172 * When 0, @ctx->attr is the found attribute and it is in mft record
1173 * @ctx->mrec. If an attribute list attribute is present, @ctx->al_entry is
1174 * the attribute list entry of the found attribute.
1175 *
1176 * When -ENOENT, @ctx->attr is the attribute which collates just after the
1177 * attribute being searched for, i.e. if one wants to add the attribute to the
1178 * mft record this is the correct place to insert it into. If an attribute
1179 * list attribute is present, @ctx->al_entry is the attribute list entry which
1180 * collates just after the attribute list entry of the attribute being searched
1181 * for, i.e. if one wants to add the attribute to the mft record this is the
1182 * correct place to insert its attribute list entry into.
1183 *
1184 * When -errno != -ENOENT, an error occurred during the lookup. @ctx->attr is
1185 * then undefined and in particular you should not rely on it not changing.
1186 */
1187int ntfs_attr_lookup(const ATTR_TYPE type, const ntfschar *name,
1188        const u32 name_len, const IGNORE_CASE_BOOL ic,
1189        const VCN lowest_vcn, const u8 *val, const u32 val_len,
1190        ntfs_attr_search_ctx *ctx)
1191{
1192    ntfs_inode *base_ni;
1193
1194    ntfs_debug("Entering.");
1195    BUG_ON(IS_ERR(ctx->mrec));
1196    if (ctx->base_ntfs_ino)
1197        base_ni = ctx->base_ntfs_ino;
1198    else
1199        base_ni = ctx->ntfs_ino;
1200    /* Sanity check, just for debugging really. */
1201    BUG_ON(!base_ni);
1202    if (!NInoAttrList(base_ni) || type == AT_ATTRIBUTE_LIST)
1203        return ntfs_attr_find(type, name, name_len, ic, val, val_len,
1204                ctx);
1205    return ntfs_external_attr_find(type, name, name_len, ic, lowest_vcn,
1206            val, val_len, ctx);
1207}
1208
1209/**
1210 * ntfs_attr_init_search_ctx - initialize an attribute search context
1211 * @ctx: attribute search context to initialize
1212 * @ni: ntfs inode with which to initialize the search context
1213 * @mrec: mft record with which to initialize the search context
1214 *
1215 * Initialize the attribute search context @ctx with @ni and @mrec.
1216 */
1217static inline void ntfs_attr_init_search_ctx(ntfs_attr_search_ctx *ctx,
1218        ntfs_inode *ni, MFT_RECORD *mrec)
1219{
1220    *ctx = (ntfs_attr_search_ctx) {
1221        .mrec = mrec,
1222        /* Sanity checks are performed elsewhere. */
1223        .attr = (ATTR_RECORD*)((u8*)mrec +
1224                le16_to_cpu(mrec->attrs_offset)),
1225        .is_first = true,
1226        .ntfs_ino = ni,
1227    };
1228}
1229
1230/**
1231 * ntfs_attr_reinit_search_ctx - reinitialize an attribute search context
1232 * @ctx: attribute search context to reinitialize
1233 *
1234 * Reinitialize the attribute search context @ctx, unmapping an associated
1235 * extent mft record if present, and initialize the search context again.
1236 *
1237 * This is used when a search for a new attribute is being started to reset
1238 * the search context to the beginning.
1239 */
1240void ntfs_attr_reinit_search_ctx(ntfs_attr_search_ctx *ctx)
1241{
1242    if (likely(!ctx->base_ntfs_ino)) {
1243        /* No attribute list. */
1244        ctx->is_first = true;
1245        /* Sanity checks are performed elsewhere. */
1246        ctx->attr = (ATTR_RECORD*)((u8*)ctx->mrec +
1247                le16_to_cpu(ctx->mrec->attrs_offset));
1248        /*
1249         * This needs resetting due to ntfs_external_attr_find() which
1250         * can leave it set despite having zeroed ctx->base_ntfs_ino.
1251         */
1252        ctx->al_entry = NULL;
1253        return;
1254    } /* Attribute list. */
1255    if (ctx->ntfs_ino != ctx->base_ntfs_ino)
1256        unmap_extent_mft_record(ctx->ntfs_ino);
1257    ntfs_attr_init_search_ctx(ctx, ctx->base_ntfs_ino, ctx->base_mrec);
1258    return;
1259}
1260
1261/**
1262 * ntfs_attr_get_search_ctx - allocate/initialize a new attribute search context
1263 * @ni: ntfs inode with which to initialize the search context
1264 * @mrec: mft record with which to initialize the search context
1265 *
1266 * Allocate a new attribute search context, initialize it with @ni and @mrec,
1267 * and return it. Return NULL if allocation failed.
1268 */
1269ntfs_attr_search_ctx *ntfs_attr_get_search_ctx(ntfs_inode *ni, MFT_RECORD *mrec)
1270{
1271    ntfs_attr_search_ctx *ctx;
1272
1273    ctx = kmem_cache_alloc(ntfs_attr_ctx_cache, GFP_NOFS);
1274    if (ctx)
1275        ntfs_attr_init_search_ctx(ctx, ni, mrec);
1276    return ctx;
1277}
1278
1279/**
1280 * ntfs_attr_put_search_ctx - release an attribute search context
1281 * @ctx: attribute search context to free
1282 *
1283 * Release the attribute search context @ctx, unmapping an associated extent
1284 * mft record if present.
1285 */
1286void ntfs_attr_put_search_ctx(ntfs_attr_search_ctx *ctx)
1287{
1288    if (ctx->base_ntfs_ino && ctx->ntfs_ino != ctx->base_ntfs_ino)
1289        unmap_extent_mft_record(ctx->ntfs_ino);
1290    kmem_cache_free(ntfs_attr_ctx_cache, ctx);
1291    return;
1292}
1293
1294#ifdef NTFS_RW
1295
1296/**
1297 * ntfs_attr_find_in_attrdef - find an attribute in the $AttrDef system file
1298 * @vol: ntfs volume to which the attribute belongs
1299 * @type: attribute type which to find
1300 *
1301 * Search for the attribute definition record corresponding to the attribute
1302 * @type in the $AttrDef system file.
1303 *
1304 * Return the attribute type definition record if found and NULL if not found.
1305 */
1306static ATTR_DEF *ntfs_attr_find_in_attrdef(const ntfs_volume *vol,
1307        const ATTR_TYPE type)
1308{
1309    ATTR_DEF *ad;
1310
1311    BUG_ON(!vol->attrdef);
1312    BUG_ON(!type);
1313    for (ad = vol->attrdef; (u8*)ad - (u8*)vol->attrdef <
1314            vol->attrdef_size && ad->type; ++ad) {
1315        /* We have not found it yet, carry on searching. */
1316        if (likely(le32_to_cpu(ad->type) < le32_to_cpu(type)))
1317            continue;
1318        /* We found the attribute; return it. */
1319        if (likely(ad->type == type))
1320            return ad;
1321        /* We have gone too far already. No point in continuing. */
1322        break;
1323    }
1324    /* Attribute not found. */
1325    ntfs_debug("Attribute type 0x%x not found in $AttrDef.",
1326            le32_to_cpu(type));
1327    return NULL;
1328}
1329
1330/**
1331 * ntfs_attr_size_bounds_check - check a size of an attribute type for validity
1332 * @vol: ntfs volume to which the attribute belongs
1333 * @type: attribute type which to check
1334 * @size: size which to check
1335 *
1336 * Check whether the @size in bytes is valid for an attribute of @type on the
1337 * ntfs volume @vol. This information is obtained from $AttrDef system file.
1338 *
1339 * Return 0 if valid, -ERANGE if not valid, or -ENOENT if the attribute is not
1340 * listed in $AttrDef.
1341 */
1342int ntfs_attr_size_bounds_check(const ntfs_volume *vol, const ATTR_TYPE type,
1343        const s64 size)
1344{
1345    ATTR_DEF *ad;
1346
1347    BUG_ON(size < 0);
1348    /*
1349     * $ATTRIBUTE_LIST has a maximum size of 256kiB, but this is not
1350     * listed in $AttrDef.
1351     */
1352    if (unlikely(type == AT_ATTRIBUTE_LIST && size > 256 * 1024))
1353        return -ERANGE;
1354    /* Get the $AttrDef entry for the attribute @type. */
1355    ad = ntfs_attr_find_in_attrdef(vol, type);
1356    if (unlikely(!ad))
1357        return -ENOENT;
1358    /* Do the bounds check. */
1359    if (((sle64_to_cpu(ad->min_size) > 0) &&
1360            size < sle64_to_cpu(ad->min_size)) ||
1361            ((sle64_to_cpu(ad->max_size) > 0) && size >
1362            sle64_to_cpu(ad->max_size)))
1363        return -ERANGE;
1364    return 0;
1365}
1366
1367/**
1368 * ntfs_attr_can_be_non_resident - check if an attribute can be non-resident
1369 * @vol: ntfs volume to which the attribute belongs
1370 * @type: attribute type which to check
1371 *
1372 * Check whether the attribute of @type on the ntfs volume @vol is allowed to
1373 * be non-resident. This information is obtained from $AttrDef system file.
1374 *
1375 * Return 0 if the attribute is allowed to be non-resident, -EPERM if not, and
1376 * -ENOENT if the attribute is not listed in $AttrDef.
1377 */
1378int ntfs_attr_can_be_non_resident(const ntfs_volume *vol, const ATTR_TYPE type)
1379{
1380    ATTR_DEF *ad;
1381
1382    /* Find the attribute definition record in $AttrDef. */
1383    ad = ntfs_attr_find_in_attrdef(vol, type);
1384    if (unlikely(!ad))
1385        return -ENOENT;
1386    /* Check the flags and return the result. */
1387    if (ad->flags & ATTR_DEF_RESIDENT)
1388        return -EPERM;
1389    return 0;
1390}
1391
1392/**
1393 * ntfs_attr_can_be_resident - check if an attribute can be resident
1394 * @vol: ntfs volume to which the attribute belongs
1395 * @type: attribute type which to check
1396 *
1397 * Check whether the attribute of @type on the ntfs volume @vol is allowed to
1398 * be resident. This information is derived from our ntfs knowledge and may
1399 * not be completely accurate, especially when user defined attributes are
1400 * present. Basically we allow everything to be resident except for index
1401 * allocation and $EA attributes.
1402 *
1403 * Return 0 if the attribute is allowed to be non-resident and -EPERM if not.
1404 *
1405 * Warning: In the system file $MFT the attribute $Bitmap must be non-resident
1406 * otherwise windows will not boot (blue screen of death)! We cannot
1407 * check for this here as we do not know which inode's $Bitmap is
1408 * being asked about so the caller needs to special case this.
1409 */
1410int ntfs_attr_can_be_resident(const ntfs_volume *vol, const ATTR_TYPE type)
1411{
1412    if (type == AT_INDEX_ALLOCATION)
1413        return -EPERM;
1414    return 0;
1415}
1416
1417/**
1418 * ntfs_attr_record_resize - resize an attribute record
1419 * @m: mft record containing attribute record
1420 * @a: attribute record to resize
1421 * @new_size: new size in bytes to which to resize the attribute record @a
1422 *
1423 * Resize the attribute record @a, i.e. the resident part of the attribute, in
1424 * the mft record @m to @new_size bytes.
1425 *
1426 * Return 0 on success and -errno on error. The following error codes are
1427 * defined:
1428 * -ENOSPC - Not enough space in the mft record @m to perform the resize.
1429 *
1430 * Note: On error, no modifications have been performed whatsoever.
1431 *
1432 * Warning: If you make a record smaller without having copied all the data you
1433 * are interested in the data may be overwritten.
1434 */
1435int ntfs_attr_record_resize(MFT_RECORD *m, ATTR_RECORD *a, u32 new_size)
1436{
1437    ntfs_debug("Entering for new_size %u.", new_size);
1438    /* Align to 8 bytes if it is not already done. */
1439    if (new_size & 7)
1440        new_size = (new_size + 7) & ~7;
1441    /* If the actual attribute length has changed, move things around. */
1442    if (new_size != le32_to_cpu(a->length)) {
1443        u32 new_muse = le32_to_cpu(m->bytes_in_use) -
1444                le32_to_cpu(a->length) + new_size;
1445        /* Not enough space in this mft record. */
1446        if (new_muse > le32_to_cpu(m->bytes_allocated))
1447            return -ENOSPC;
1448        /* Move attributes following @a to their new location. */
1449        memmove((u8*)a + new_size, (u8*)a + le32_to_cpu(a->length),
1450                le32_to_cpu(m->bytes_in_use) - ((u8*)a -
1451                (u8*)m) - le32_to_cpu(a->length));
1452        /* Adjust @m to reflect the change in used space. */
1453        m->bytes_in_use = cpu_to_le32(new_muse);
1454        /* Adjust @a to reflect the new size. */
1455        if (new_size >= offsetof(ATTR_REC, length) + sizeof(a->length))
1456            a->length = cpu_to_le32(new_size);
1457    }
1458    return 0;
1459}
1460
1461/**
1462 * ntfs_resident_attr_value_resize - resize the value of a resident attribute
1463 * @m: mft record containing attribute record
1464 * @a: attribute record whose value to resize
1465 * @new_size: new size in bytes to which to resize the attribute value of @a
1466 *
1467 * Resize the value of the attribute @a in the mft record @m to @new_size bytes.
1468 * If the value is made bigger, the newly allocated space is cleared.
1469 *
1470 * Return 0 on success and -errno on error. The following error codes are
1471 * defined:
1472 * -ENOSPC - Not enough space in the mft record @m to perform the resize.
1473 *
1474 * Note: On error, no modifications have been performed whatsoever.
1475 *
1476 * Warning: If you make a record smaller without having copied all the data you
1477 * are interested in the data may be overwritten.
1478 */
1479int ntfs_resident_attr_value_resize(MFT_RECORD *m, ATTR_RECORD *a,
1480        const u32 new_size)
1481{
1482    u32 old_size;
1483
1484    /* Resize the resident part of the attribute record. */
1485    if (ntfs_attr_record_resize(m, a,
1486            le16_to_cpu(a->data.resident.value_offset) + new_size))
1487        return -ENOSPC;
1488    /*
1489     * The resize succeeded! If we made the attribute value bigger, clear
1490     * the area between the old size and @new_size.
1491     */
1492    old_size = le32_to_cpu(a->data.resident.value_length);
1493    if (new_size > old_size)
1494        memset((u8*)a + le16_to_cpu(a->data.resident.value_offset) +
1495                old_size, 0, new_size - old_size);
1496    /* Finally update the length of the attribute value. */
1497    a->data.resident.value_length = cpu_to_le32(new_size);
1498    return 0;
1499}
1500
1501/**
1502 * ntfs_attr_make_non_resident - convert a resident to a non-resident attribute
1503 * @ni: ntfs inode describing the attribute to convert
1504 * @data_size: size of the resident data to copy to the non-resident attribute
1505 *
1506 * Convert the resident ntfs attribute described by the ntfs inode @ni to a
1507 * non-resident one.
1508 *
1509 * @data_size must be equal to the attribute value size. This is needed since
1510 * we need to know the size before we can map the mft record and our callers
1511 * always know it. The reason we cannot simply read the size from the vfs
1512 * inode i_size is that this is not necessarily uptodate. This happens when
1513 * ntfs_attr_make_non_resident() is called in the ->truncate call path(s).
1514 *
1515 * Return 0 on success and -errno on error. The following error return codes
1516 * are defined:
1517 * -EPERM - The attribute is not allowed to be non-resident.
1518 * -ENOMEM - Not enough memory.
1519 * -ENOSPC - Not enough disk space.
1520 * -EINVAL - Attribute not defined on the volume.
1521 * -EIO - I/o error or other error.
1522 * Note that -ENOSPC is also returned in the case that there is not enough
1523 * space in the mft record to do the conversion. This can happen when the mft
1524 * record is already very full. The caller is responsible for trying to make
1525 * space in the mft record and trying again. FIXME: Do we need a separate
1526 * error return code for this kind of -ENOSPC or is it always worth trying
1527 * again in case the attribute may then fit in a resident state so no need to
1528 * make it non-resident at all? Ho-hum... (AIA)
1529 *
1530 * NOTE to self: No changes in the attribute list are required to move from
1531 * a resident to a non-resident attribute.
1532 *
1533 * Locking: - The caller must hold i_mutex on the inode.
1534 */
1535int ntfs_attr_make_non_resident(ntfs_inode *ni, const u32 data_size)
1536{
1537    s64 new_size;
1538    struct inode *vi = VFS_I(ni);
1539    ntfs_volume *vol = ni->vol;
1540    ntfs_inode *base_ni;
1541    MFT_RECORD *m;
1542    ATTR_RECORD *a;
1543    ntfs_attr_search_ctx *ctx;
1544    struct page *page;
1545    runlist_element *rl;
1546    u8 *kaddr;
1547    unsigned long flags;
1548    int mp_size, mp_ofs, name_ofs, arec_size, err, err2;
1549    u32 attr_size;
1550    u8 old_res_attr_flags;
1551
1552    /* Check that the attribute is allowed to be non-resident. */
1553    err = ntfs_attr_can_be_non_resident(vol, ni->type);
1554    if (unlikely(err)) {
1555        if (err == -EPERM)
1556            ntfs_debug("Attribute is not allowed to be "
1557                    "non-resident.");
1558        else
1559            ntfs_debug("Attribute not defined on the NTFS "
1560                    "volume!");
1561        return err;
1562    }
1563    /*
1564     * FIXME: Compressed and encrypted attributes are not supported when
1565     * writing and we should never have gotten here for them.
1566     */
1567    BUG_ON(NInoCompressed(ni));
1568    BUG_ON(NInoEncrypted(ni));
1569    /*
1570     * The size needs to be aligned to a cluster boundary for allocation
1571     * purposes.
1572     */
1573    new_size = (data_size + vol->cluster_size - 1) &
1574            ~(vol->cluster_size - 1);
1575    if (new_size > 0) {
1576        /*
1577         * Will need the page later and since the page lock nests
1578         * outside all ntfs locks, we need to get the page now.
1579         */
1580        page = find_or_create_page(vi->i_mapping, 0,
1581                mapping_gfp_mask(vi->i_mapping));
1582        if (unlikely(!page))
1583            return -ENOMEM;
1584        /* Start by allocating clusters to hold the attribute value. */
1585        rl = ntfs_cluster_alloc(vol, 0, new_size >>
1586                vol->cluster_size_bits, -1, DATA_ZONE, true);
1587        if (IS_ERR(rl)) {
1588            err = PTR_ERR(rl);
1589            ntfs_debug("Failed to allocate cluster%s, error code "
1590                    "%i.", (new_size >>
1591                    vol->cluster_size_bits) > 1 ? "s" : "",
1592                    err);
1593            goto page_err_out;
1594        }
1595    } else {
1596        rl = NULL;
1597        page = NULL;
1598    }
1599    /* Determine the size of the mapping pairs array. */
1600    mp_size = ntfs_get_size_for_mapping_pairs(vol, rl, 0, -1);
1601    if (unlikely(mp_size < 0)) {
1602        err = mp_size;
1603        ntfs_debug("Failed to get size for mapping pairs array, error "
1604                "code %i.", err);
1605        goto rl_err_out;
1606    }
1607    down_write(&ni->runlist.lock);
1608    if (!NInoAttr(ni))
1609        base_ni = ni;
1610    else
1611        base_ni = ni->ext.base_ntfs_ino;
1612    m = map_mft_record(base_ni);
1613    if (IS_ERR(m)) {
1614        err = PTR_ERR(m);
1615        m = NULL;
1616        ctx = NULL;
1617        goto err_out;
1618    }
1619    ctx = ntfs_attr_get_search_ctx(base_ni, m);
1620    if (unlikely(!ctx)) {
1621        err = -ENOMEM;
1622        goto err_out;
1623    }
1624    err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
1625            CASE_SENSITIVE, 0, NULL, 0, ctx);
1626    if (unlikely(err)) {
1627        if (err == -ENOENT)
1628            err = -EIO;
1629        goto err_out;
1630    }
1631    m = ctx->mrec;
1632    a = ctx->attr;
1633    BUG_ON(NInoNonResident(ni));
1634    BUG_ON(a->non_resident);
1635    /*
1636     * Calculate new offsets for the name and the mapping pairs array.
1637     */
1638    if (NInoSparse(ni) || NInoCompressed(ni))
1639        name_ofs = (offsetof(ATTR_REC,
1640                data.non_resident.compressed_size) +
1641                sizeof(a->data.non_resident.compressed_size) +
1642                7) & ~7;
1643    else
1644        name_ofs = (offsetof(ATTR_REC,
1645                data.non_resident.compressed_size) + 7) & ~7;
1646    mp_ofs = (name_ofs + a->name_length * sizeof(ntfschar) + 7) & ~7;
1647    /*
1648     * Determine the size of the resident part of the now non-resident
1649     * attribute record.
1650     */
1651    arec_size = (mp_ofs + mp_size + 7) & ~7;
1652    /*
1653     * If the page is not uptodate bring it uptodate by copying from the
1654     * attribute value.
1655     */
1656    attr_size = le32_to_cpu(a->data.resident.value_length);
1657    BUG_ON(attr_size != data_size);
1658    if (page && !PageUptodate(page)) {
1659        kaddr = kmap_atomic(page, KM_USER0);
1660        memcpy(kaddr, (u8*)a +
1661                le16_to_cpu(a->data.resident.value_offset),
1662                attr_size);
1663        memset(kaddr + attr_size, 0, PAGE_CACHE_SIZE - attr_size);
1664        kunmap_atomic(kaddr, KM_USER0);
1665        flush_dcache_page(page);
1666        SetPageUptodate(page);
1667    }
1668    /* Backup the attribute flag. */
1669    old_res_attr_flags = a->data.resident.flags;
1670    /* Resize the resident part of the attribute record. */
1671    err = ntfs_attr_record_resize(m, a, arec_size);
1672    if (unlikely(err))
1673        goto err_out;
1674    /*
1675     * Convert the resident part of the attribute record to describe a
1676     * non-resident attribute.
1677     */
1678    a->non_resident = 1;
1679    /* Move the attribute name if it exists and update the offset. */
1680    if (a->name_length)
1681        memmove((u8*)a + name_ofs, (u8*)a + le16_to_cpu(a->name_offset),
1682                a->name_length * sizeof(ntfschar));
1683    a->name_offset = cpu_to_le16(name_ofs);
1684    /* Setup the fields specific to non-resident attributes. */
1685    a->data.non_resident.lowest_vcn = 0;
1686    a->data.non_resident.highest_vcn = cpu_to_sle64((new_size - 1) >>
1687            vol->cluster_size_bits);
1688    a->data.non_resident.mapping_pairs_offset = cpu_to_le16(mp_ofs);
1689    memset(&a->data.non_resident.reserved, 0,
1690            sizeof(a->data.non_resident.reserved));
1691    a->data.non_resident.allocated_size = cpu_to_sle64(new_size);
1692    a->data.non_resident.data_size =
1693            a->data.non_resident.initialized_size =
1694            cpu_to_sle64(attr_size);
1695    if (NInoSparse(ni) || NInoCompressed(ni)) {
1696        a->data.non_resident.compression_unit = 0;
1697        if (NInoCompressed(ni) || vol->major_ver < 3)
1698            a->data.non_resident.compression_unit = 4;
1699        a->data.non_resident.compressed_size =
1700                a->data.non_resident.allocated_size;
1701    } else
1702        a->data.non_resident.compression_unit = 0;
1703    /* Generate the mapping pairs array into the attribute record. */
1704    err = ntfs_mapping_pairs_build(vol, (u8*)a + mp_ofs,
1705            arec_size - mp_ofs, rl, 0, -1, NULL);
1706    if (unlikely(err)) {
1707        ntfs_debug("Failed to build mapping pairs, error code %i.",
1708                err);
1709        goto undo_err_out;
1710    }
1711    /* Setup the in-memory attribute structure to be non-resident. */
1712    ni->runlist.rl = rl;
1713    write_lock_irqsave(&ni->size_lock, flags);
1714    ni->allocated_size = new_size;
1715    if (NInoSparse(ni) || NInoCompressed(ni)) {
1716        ni->itype.compressed.size = ni->allocated_size;
1717        if (a->data.non_resident.compression_unit) {
1718            ni->itype.compressed.block_size = 1U << (a->data.
1719                    non_resident.compression_unit +
1720                    vol->cluster_size_bits);
1721            ni->itype.compressed.block_size_bits =
1722                    ffs(ni->itype.compressed.block_size) -
1723                    1;
1724            ni->itype.compressed.block_clusters = 1U <<
1725                    a->data.non_resident.compression_unit;
1726        } else {
1727            ni->itype.compressed.block_size = 0;
1728            ni->itype.compressed.block_size_bits = 0;
1729            ni->itype.compressed.block_clusters = 0;
1730        }
1731        vi->i_blocks = ni->itype.compressed.size >> 9;
1732    } else
1733        vi->i_blocks = ni->allocated_size >> 9;
1734    write_unlock_irqrestore(&ni->size_lock, flags);
1735    /*
1736     * This needs to be last since the address space operations ->readpage
1737     * and ->writepage can run concurrently with us as they are not
1738     * serialized on i_mutex. Note, we are not allowed to fail once we flip
1739     * this switch, which is another reason to do this last.
1740     */
1741    NInoSetNonResident(ni);
1742    /* Mark the mft record dirty, so it gets written back. */
1743    flush_dcache_mft_record_page(ctx->ntfs_ino);
1744    mark_mft_record_dirty(ctx->ntfs_ino);
1745    ntfs_attr_put_search_ctx(ctx);
1746    unmap_mft_record(base_ni);
1747    up_write(&ni->runlist.lock);
1748    if (page) {
1749        set_page_dirty(page);
1750        unlock_page(page);
1751        mark_page_accessed(page);
1752        page_cache_release(page);
1753    }
1754    ntfs_debug("Done.");
1755    return 0;
1756undo_err_out:
1757    /* Convert the attribute back into a resident attribute. */
1758    a->non_resident = 0;
1759    /* Move the attribute name if it exists and update the offset. */
1760    name_ofs = (offsetof(ATTR_RECORD, data.resident.reserved) +
1761            sizeof(a->data.resident.reserved) + 7) & ~7;
1762    if (a->name_length)
1763        memmove((u8*)a + name_ofs, (u8*)a + le16_to_cpu(a->name_offset),
1764                a->name_length * sizeof(ntfschar));
1765    mp_ofs = (name_ofs + a->name_length * sizeof(ntfschar) + 7) & ~7;
1766    a->name_offset = cpu_to_le16(name_ofs);
1767    arec_size = (mp_ofs + attr_size + 7) & ~7;
1768    /* Resize the resident part of the attribute record. */
1769    err2 = ntfs_attr_record_resize(m, a, arec_size);
1770    if (unlikely(err2)) {
1771        /*
1772         * This cannot happen (well if memory corruption is at work it
1773         * could happen in theory), but deal with it as well as we can.
1774         * If the old size is too small, truncate the attribute,
1775         * otherwise simply give it a larger allocated size.
1776         * FIXME: Should check whether chkdsk complains when the
1777         * allocated size is much bigger than the resident value size.
1778         */
1779        arec_size = le32_to_cpu(a->length);
1780        if ((mp_ofs + attr_size) > arec_size) {
1781            err2 = attr_size;
1782            attr_size = arec_size - mp_ofs;
1783            ntfs_error(vol->sb, "Failed to undo partial resident "
1784                    "to non-resident attribute "
1785                    "conversion. Truncating inode 0x%lx, "
1786                    "attribute type 0x%x from %i bytes to "
1787                    "%i bytes to maintain metadata "
1788                    "consistency. THIS MEANS YOU ARE "
1789                    "LOSING %i BYTES DATA FROM THIS %s.",
1790                    vi->i_ino,
1791                    (unsigned)le32_to_cpu(ni->type),
1792                    err2, attr_size, err2 - attr_size,
1793                    ((ni->type == AT_DATA) &&
1794                    !ni->name_len) ? "FILE": "ATTRIBUTE");
1795            write_lock_irqsave(&ni->size_lock, flags);
1796            ni->initialized_size = attr_size;
1797            i_size_write(vi, attr_size);
1798            write_unlock_irqrestore(&ni->size_lock, flags);
1799        }
1800    }
1801    /* Setup the fields specific to resident attributes. */
1802    a->data.resident.value_length = cpu_to_le32(attr_size);
1803    a->data.resident.value_offset = cpu_to_le16(mp_ofs);
1804    a->data.resident.flags = old_res_attr_flags;
1805    memset(&a->data.resident.reserved, 0,
1806            sizeof(a->data.resident.reserved));
1807    /* Copy the data from the page back to the attribute value. */
1808    if (page) {
1809        kaddr = kmap_atomic(page, KM_USER0);
1810        memcpy((u8*)a + mp_ofs, kaddr, attr_size);
1811        kunmap_atomic(kaddr, KM_USER0);
1812    }
1813    /* Setup the allocated size in the ntfs inode in case it changed. */
1814    write_lock_irqsave(&ni->size_lock, flags);
1815    ni->allocated_size = arec_size - mp_ofs;
1816    write_unlock_irqrestore(&ni->size_lock, flags);
1817    /* Mark the mft record dirty, so it gets written back. */
1818    flush_dcache_mft_record_page(ctx->ntfs_ino);
1819    mark_mft_record_dirty(ctx->ntfs_ino);
1820err_out:
1821    if (ctx)
1822        ntfs_attr_put_search_ctx(ctx);
1823    if (m)
1824        unmap_mft_record(base_ni);
1825    ni->runlist.rl = NULL;
1826    up_write(&ni->runlist.lock);
1827rl_err_out:
1828    if (rl) {
1829        if (ntfs_cluster_free_from_rl(vol, rl) < 0) {
1830            ntfs_error(vol->sb, "Failed to release allocated "
1831                    "cluster(s) in error code path. Run "
1832                    "chkdsk to recover the lost "
1833                    "cluster(s).");
1834            NVolSetErrors(vol);
1835        }
1836        ntfs_free(rl);
1837page_err_out:
1838        unlock_page(page);
1839        page_cache_release(page);
1840    }
1841    if (err == -EINVAL)
1842        err = -EIO;
1843    return err;
1844}
1845
1846/**
1847 * ntfs_attr_extend_allocation - extend the allocated space of an attribute
1848 * @ni: ntfs inode of the attribute whose allocation to extend
1849 * @new_alloc_size: new size in bytes to which to extend the allocation to
1850 * @new_data_size: new size in bytes to which to extend the data to
1851 * @data_start: beginning of region which is required to be non-sparse
1852 *
1853 * Extend the allocated space of an attribute described by the ntfs inode @ni
1854 * to @new_alloc_size bytes. If @data_start is -1, the whole extension may be
1855 * implemented as a hole in the file (as long as both the volume and the ntfs
1856 * inode @ni have sparse support enabled). If @data_start is >= 0, then the
1857 * region between the old allocated size and @data_start - 1 may be made sparse
1858 * but the regions between @data_start and @new_alloc_size must be backed by
1859 * actual clusters.
1860 *
1861 * If @new_data_size is -1, it is ignored. If it is >= 0, then the data size
1862 * of the attribute is extended to @new_data_size. Note that the i_size of the
1863 * vfs inode is not updated. Only the data size in the base attribute record
1864 * is updated. The caller has to update i_size separately if this is required.
1865 * WARNING: It is a BUG() for @new_data_size to be smaller than the old data
1866 * size as well as for @new_data_size to be greater than @new_alloc_size.
1867 *
1868 * For resident attributes this involves resizing the attribute record and if
1869 * necessary moving it and/or other attributes into extent mft records and/or
1870 * converting the attribute to a non-resident attribute which in turn involves
1871 * extending the allocation of a non-resident attribute as described below.
1872 *
1873 * For non-resident attributes this involves allocating clusters in the data
1874 * zone on the volume (except for regions that are being made sparse) and
1875 * extending the run list to describe the allocated clusters as well as
1876 * updating the mapping pairs array of the attribute. This in turn involves
1877 * resizing the attribute record and if necessary moving it and/or other
1878 * attributes into extent mft records and/or splitting the attribute record
1879 * into multiple extent attribute records.
1880 *
1881 * Also, the attribute list attribute is updated if present and in some of the
1882 * above cases (the ones where extent mft records/attributes come into play),
1883 * an attribute list attribute is created if not already present.
1884 *
1885 * Return the new allocated size on success and -errno on error. In the case
1886 * that an error is encountered but a partial extension at least up to
1887 * @data_start (if present) is possible, the allocation is partially extended
1888 * and this is returned. This means the caller must check the returned size to
1889 * determine if the extension was partial. If @data_start is -1 then partial
1890 * allocations are not performed.
1891 *
1892 * WARNING: Do not call ntfs_attr_extend_allocation() for $MFT/$DATA.
1893 *
1894 * Locking: This function takes the runlist lock of @ni for writing as well as
1895 * locking the mft record of the base ntfs inode. These locks are maintained
1896 * throughout execution of the function. These locks are required so that the
1897 * attribute can be resized safely and so that it can for example be converted
1898 * from resident to non-resident safely.
1899 *
1900 * TODO: At present attribute list attribute handling is not implemented.
1901 *
1902 * TODO: At present it is not safe to call this function for anything other
1903 * than the $DATA attribute(s) of an uncompressed and unencrypted file.
1904 */
1905s64 ntfs_attr_extend_allocation(ntfs_inode *ni, s64 new_alloc_size,
1906        const s64 new_data_size, const s64 data_start)
1907{
1908    VCN vcn;
1909    s64 ll, allocated_size, start = data_start;
1910    struct inode *vi = VFS_I(ni);
1911    ntfs_volume *vol = ni->vol;
1912    ntfs_inode *base_ni;
1913    MFT_RECORD *m;
1914    ATTR_RECORD *a;
1915    ntfs_attr_search_ctx *ctx;
1916    runlist_element *rl, *rl2;
1917    unsigned long flags;
1918    int err, mp_size;
1919    u32 attr_len = 0; /* Silence stupid gcc warning. */
1920    bool mp_rebuilt;
1921
1922#ifdef DEBUG
1923    read_lock_irqsave(&ni->size_lock, flags);
1924    allocated_size = ni->allocated_size;
1925    read_unlock_irqrestore(&ni->size_lock, flags);
1926    ntfs_debug("Entering for i_ino 0x%lx, attribute type 0x%x, "
1927            "old_allocated_size 0x%llx, "
1928            "new_allocated_size 0x%llx, new_data_size 0x%llx, "
1929            "data_start 0x%llx.", vi->i_ino,
1930            (unsigned)le32_to_cpu(ni->type),
1931            (unsigned long long)allocated_size,
1932            (unsigned long long)new_alloc_size,
1933            (unsigned long long)new_data_size,
1934            (unsigned long long)start);
1935#endif
1936retry_extend:
1937    /*
1938     * For non-resident attributes, @start and @new_size need to be aligned
1939     * to cluster boundaries for allocation purposes.
1940     */
1941    if (NInoNonResident(ni)) {
1942        if (start > 0)
1943            start &= ~(s64)vol->cluster_size_mask;
1944        new_alloc_size = (new_alloc_size + vol->cluster_size - 1) &
1945                ~(s64)vol->cluster_size_mask;
1946    }
1947    BUG_ON(new_data_size >= 0 && new_data_size > new_alloc_size);
1948    /* Check if new size is allowed in $AttrDef. */
1949    err = ntfs_attr_size_bounds_check(vol, ni->type, new_alloc_size);
1950    if (unlikely(err)) {
1951        /* Only emit errors when the write will fail completely. */
1952        read_lock_irqsave(&ni->size_lock, flags);
1953        allocated_size = ni->allocated_size;
1954        read_unlock_irqrestore(&ni->size_lock, flags);
1955        if (start < 0 || start >= allocated_size) {
1956            if (err == -ERANGE) {
1957                ntfs_error(vol->sb, "Cannot extend allocation "
1958                        "of inode 0x%lx, attribute "
1959                        "type 0x%x, because the new "
1960                        "allocation would exceed the "
1961                        "maximum allowed size for "
1962                        "this attribute type.",
1963                        vi->i_ino, (unsigned)
1964                        le32_to_cpu(ni->type));
1965            } else {
1966                ntfs_error(vol->sb, "Cannot extend allocation "
1967                        "of inode 0x%lx, attribute "
1968                        "type 0x%x, because this "
1969                        "attribute type is not "
1970                        "defined on the NTFS volume. "
1971                        "Possible corruption! You "
1972                        "should run chkdsk!",
1973                        vi->i_ino, (unsigned)
1974                        le32_to_cpu(ni->type));
1975            }
1976        }
1977        /* Translate error code to be POSIX conformant for write(2). */
1978        if (err == -ERANGE)
1979            err = -EFBIG;
1980        else
1981            err = -EIO;
1982        return err;
1983    }
1984    if (!NInoAttr(ni))
1985        base_ni = ni;
1986    else
1987        base_ni = ni->ext.base_ntfs_ino;
1988    /*
1989     * We will be modifying both the runlist (if non-resident) and the mft
1990     * record so lock them both down.
1991     */
1992    down_write(&ni->runlist.lock);
1993    m = map_mft_record(base_ni);
1994    if (IS_ERR(m)) {
1995        err = PTR_ERR(m);
1996        m = NULL;
1997        ctx = NULL;
1998        goto err_out;
1999    }
2000    ctx = ntfs_attr_get_search_ctx(base_ni, m);
2001    if (unlikely(!ctx)) {
2002        err = -ENOMEM;
2003        goto err_out;
2004    }
2005    read_lock_irqsave(&ni->size_lock, flags);
2006    allocated_size = ni->allocated_size;
2007    read_unlock_irqrestore(&ni->size_lock, flags);
2008    /*
2009     * If non-resident, seek to the last extent. If resident, there is
2010     * only one extent, so seek to that.
2011     */
2012    vcn = NInoNonResident(ni) ? allocated_size >> vol->cluster_size_bits :
2013            0;
2014    /*
2015     * Abort if someone did the work whilst we waited for the locks. If we
2016     * just converted the attribute from resident to non-resident it is
2017     * likely that exactly this has happened already. We cannot quite
2018     * abort if we need to update the data size.
2019     */
2020    if (unlikely(new_alloc_size <= allocated_size)) {
2021        ntfs_debug("Allocated size already exceeds requested size.");
2022        new_alloc_size = allocated_size;
2023        if (new_data_size < 0)
2024            goto done;
2025        /*
2026         * We want the first attribute extent so that we can update the
2027         * data size.
2028         */
2029        vcn = 0;
2030    }
2031    err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
2032            CASE_SENSITIVE, vcn, NULL, 0, ctx);
2033    if (unlikely(err)) {
2034        if (err == -ENOENT)
2035            err = -EIO;
2036        goto err_out;
2037    }
2038    m = ctx->mrec;
2039    a = ctx->attr;
2040    /* Use goto to reduce indentation. */
2041    if (a->non_resident)
2042        goto do_non_resident_extend;
2043    BUG_ON(NInoNonResident(ni));
2044    /* The total length of the attribute value. */
2045    attr_len = le32_to_cpu(a->data.resident.value_length);
2046    /*
2047     * Extend the attribute record to be able to store the new attribute
2048     * size. ntfs_attr_record_resize() will not do anything if the size is
2049     * not changing.
2050     */
2051    if (new_alloc_size < vol->mft_record_size &&
2052            !ntfs_attr_record_resize(m, a,
2053            le16_to_cpu(a->data.resident.value_offset) +
2054            new_alloc_size)) {
2055        /* The resize succeeded! */
2056        write_lock_irqsave(&ni->size_lock, flags);
2057        ni->allocated_size = le32_to_cpu(a->length) -
2058                le16_to_cpu(a->data.resident.value_offset);
2059        write_unlock_irqrestore(&ni->size_lock, flags);
2060        if (new_data_size >= 0) {
2061            BUG_ON(new_data_size < attr_len);
2062            a->data.resident.value_length =
2063                    cpu_to_le32((u32)new_data_size);
2064        }
2065        goto flush_done;
2066    }
2067    /*
2068     * We have to drop all the locks so we can call
2069     * ntfs_attr_make_non_resident(). This could be optimised by try-
2070     * locking the first page cache page and only if that fails dropping
2071     * the locks, locking the page, and redoing all the locking and
2072     * lookups. While this would be a huge optimisation, it is not worth
2073     * it as this is definitely a slow code path.
2074     */
2075    ntfs_attr_put_search_ctx(ctx);
2076    unmap_mft_record(base_ni);
2077    up_write(&ni->runlist.lock);
2078    /*
2079     * Not enough space in the mft record, try to make the attribute
2080     * non-resident and if successful restart the extension process.
2081     */
2082    err = ntfs_attr_make_non_resident(ni, attr_len);
2083    if (likely(!err))
2084        goto retry_extend;
2085    /*
2086     * Could not make non-resident. If this is due to this not being
2087     * permitted for this attribute type or there not being enough space,
2088     * try to make other attributes non-resident. Otherwise fail.
2089     */
2090    if (unlikely(err != -EPERM && err != -ENOSPC)) {
2091        /* Only emit errors when the write will fail completely. */
2092        read_lock_irqsave(&ni->size_lock, flags);
2093        allocated_size = ni->allocated_size;
2094        read_unlock_irqrestore(&ni->size_lock, flags);
2095        if (start < 0 || start >= allocated_size)
2096            ntfs_error(vol->sb, "Cannot extend allocation of "
2097                    "inode 0x%lx, attribute type 0x%x, "
2098                    "because the conversion from resident "
2099                    "to non-resident attribute failed "
2100                    "with error code %i.", vi->i_ino,
2101                    (unsigned)le32_to_cpu(ni->type), err);
2102        if (err != -ENOMEM)
2103            err = -EIO;
2104        goto conv_err_out;
2105    }
2106    /* TODO: Not implemented from here, abort. */
2107    read_lock_irqsave(&ni->size_lock, flags);
2108    allocated_size = ni->allocated_size;
2109    read_unlock_irqrestore(&ni->size_lock, flags);
2110    if (start < 0 || start >= allocated_size) {
2111        if (err == -ENOSPC)
2112            ntfs_error(vol->sb, "Not enough space in the mft "
2113                    "record/on disk for the non-resident "
2114                    "attribute value. This case is not "
2115                    "implemented yet.");
2116        else /* if (err == -EPERM) */
2117            ntfs_error(vol->sb, "This attribute type may not be "
2118                    "non-resident. This case is not "
2119                    "implemented yet.");
2120    }
2121    err = -EOPNOTSUPP;
2122    goto conv_err_out;
2123#if 0
2124    // TODO: Attempt to make other attributes non-resident.
2125    if (!err)
2126        goto do_resident_extend;
2127    /*
2128     * Both the attribute list attribute and the standard information
2129     * attribute must remain in the base inode. Thus, if this is one of
2130     * these attributes, we have to try to move other attributes out into
2131     * extent mft records instead.
2132     */
2133    if (ni->type == AT_ATTRIBUTE_LIST ||
2134            ni->type == AT_STANDARD_INFORMATION) {
2135        // TODO: Attempt to move other attributes into extent mft
2136        // records.
2137        err = -EOPNOTSUPP;
2138        if (!err)
2139            goto do_resident_extend;
2140        goto err_out;
2141    }
2142    // TODO: Attempt to move this attribute to an extent mft record, but
2143    // only if it is not already the only attribute in an mft record in
2144    // which case there would be nothing to gain.
2145    err = -EOPNOTSUPP;
2146    if (!err)
2147        goto do_resident_extend;
2148    /* There is nothing we can do to make enough space. )-: */
2149    goto err_out;
2150#endif
2151do_non_resident_extend:
2152    BUG_ON(!NInoNonResident(ni));
2153    if (new_alloc_size == allocated_size) {
2154        BUG_ON(vcn);
2155        goto alloc_done;
2156    }
2157    /*
2158     * If the data starts after the end of the old allocation, this is a
2159     * $DATA attribute and sparse attributes are enabled on the volume and
2160     * for this inode, then create a sparse region between the old
2161     * allocated size and the start of the data. Otherwise simply proceed
2162     * with filling the whole space between the old allocated size and the
2163     * new allocated size with clusters.
2164     */
2165    if ((start >= 0 && start <= allocated_size) || ni->type != AT_DATA ||
2166            !NVolSparseEnabled(vol) || NInoSparseDisabled(ni))
2167        goto skip_sparse;
2168    // TODO: This is not implemented yet. We just fill in with real
2169    // clusters for now...
2170    ntfs_debug("Inserting holes is not-implemented yet. Falling back to "
2171            "allocating real clusters instead.");
2172skip_sparse:
2173    rl = ni->runlist.rl;
2174    if (likely(rl)) {
2175        /* Seek to the end of the runlist. */
2176        while (rl->length)
2177            rl++;
2178    }
2179    /* If this attribute extent is not mapped, map it now. */
2180    if (unlikely(!rl || rl->lcn == LCN_RL_NOT_MAPPED ||
2181            (rl->lcn == LCN_ENOENT && rl > ni->runlist.rl &&
2182            (rl-1)->lcn == LCN_RL_NOT_MAPPED))) {
2183        if (!rl && !allocated_size)
2184            goto first_alloc;
2185        rl = ntfs_mapping_pairs_decompress(vol, a, ni->runlist.rl);
2186        if (IS_ERR(rl)) {
2187            err = PTR_ERR(rl);
2188            if (start < 0 || start >= allocated_size)
2189                ntfs_error(vol->sb, "Cannot extend allocation "
2190                        "of inode 0x%lx, attribute "
2191                        "type 0x%x, because the "
2192                        "mapping of a runlist "
2193                        "fragment failed with error "
2194                        "code %i.", vi->i_ino,
2195                        (unsigned)le32_to_cpu(ni->type),
2196                        err);
2197            if (err != -ENOMEM)
2198                err = -EIO;
2199            goto err_out;
2200        }
2201        ni->runlist.rl = rl;
2202        /* Seek to the end of the runlist. */
2203        while (rl->length)
2204            rl++;
2205    }
2206    /*
2207     * We now know the runlist of the last extent is mapped and @rl is at
2208     * the end of the runlist. We want to begin allocating clusters
2209     * starting at the last allocated cluster to reduce fragmentation. If
2210     * there are no valid LCNs in the attribute we let the cluster
2211     * allocator choose the starting cluster.
2212     */
2213    /* If the last LCN is a hole or simillar seek back to last real LCN. */
2214    while (rl->lcn < 0 && rl > ni->runlist.rl)
2215        rl--;
2216first_alloc:
2217    // FIXME: Need to implement partial allocations so at least part of the
2218    // write can be performed when start >= 0. (Needed for POSIX write(2)
2219    // conformance.)
2220    rl2 = ntfs_cluster_alloc(vol, allocated_size >> vol->cluster_size_bits,
2221            (new_alloc_size - allocated_size) >>
2222            vol->cluster_size_bits, (rl && (rl->lcn >= 0)) ?
2223            rl->lcn + rl->length : -1, DATA_ZONE, true);
2224    if (IS_ERR(rl2)) {
2225        err = PTR_ERR(rl2);
2226        if (start < 0 || start >= allocated_size)
2227            ntfs_error(vol->sb, "Cannot extend allocation of "
2228                    "inode 0x%lx, attribute type 0x%x, "
2229                    "because the allocation of clusters "
2230                    "failed with error code %i.", vi->i_ino,
2231                    (unsigned)le32_to_cpu(ni->type), err);
2232        if (err != -ENOMEM && err != -ENOSPC)
2233            err = -EIO;
2234        goto err_out;
2235    }
2236    rl = ntfs_runlists_merge(ni->runlist.rl, rl2);
2237    if (IS_ERR(rl)) {
2238        err = PTR_ERR(rl);
2239        if (start < 0 || start >= allocated_size)
2240            ntfs_error(vol->sb, "Cannot extend allocation of "
2241                    "inode 0x%lx, attribute type 0x%x, "
2242                    "because the runlist merge failed "
2243                    "with error code %i.", vi->i_ino,
2244                    (unsigned)le32_to_cpu(ni->type), err);
2245        if (err != -ENOMEM)
2246            err = -EIO;
2247        if (ntfs_cluster_free_from_rl(vol, rl2)) {
2248            ntfs_error(vol->sb, "Failed to release allocated "
2249                    "cluster(s) in error code path. Run "
2250                    "chkdsk to recover the lost "
2251                    "cluster(s).");
2252            NVolSetErrors(vol);
2253        }
2254        ntfs_free(rl2);
2255        goto err_out;
2256    }
2257    ni->runlist.rl = rl;
2258    ntfs_debug("Allocated 0x%llx clusters.", (long long)(new_alloc_size -
2259            allocated_size) >> vol->cluster_size_bits);
2260    /* Find the runlist element with which the attribute extent starts. */
2261    ll = sle64_to_cpu(a->data.non_resident.lowest_vcn);
2262    rl2 = ntfs_rl_find_vcn_nolock(rl, ll);
2263    BUG_ON(!rl2);
2264    BUG_ON(!rl2->length);
2265    BUG_ON(rl2->lcn < LCN_HOLE);
2266    mp_rebuilt = false;
2267    /* Get the size for the new mapping pairs array for this extent. */
2268    mp_size = ntfs_get_size_for_mapping_pairs(vol, rl2, ll, -1);
2269    if (unlikely(mp_size <= 0)) {
2270        err = mp_size;
2271        if (start < 0 || start >= allocated_size)
2272            ntfs_error(vol->sb, "Cannot extend allocation of "
2273                    "inode 0x%lx, attribute type 0x%x, "
2274                    "because determining the size for the "
2275                    "mapping pairs failed with error code "
2276                    "%i.", vi->i_ino,
2277                    (unsigned)le32_to_cpu(ni->type), err);
2278        err = -EIO;
2279        goto undo_alloc;
2280    }
2281    /* Extend the attribute record to fit the bigger mapping pairs array. */
2282    attr_len = le32_to_cpu(a->length);
2283    err = ntfs_attr_record_resize(m, a, mp_size +
2284            le16_to_cpu(a->data.non_resident.mapping_pairs_offset));
2285    if (unlikely(err)) {
2286        BUG_ON(err != -ENOSPC);
2287        // TODO: Deal with this by moving this extent to a new mft
2288        // record or by starting a new extent in a new mft record,
2289        // possibly by extending this extent partially and filling it
2290        // and creating a new extent for the remainder, or by making
2291        // other attributes non-resident and/or by moving other
2292        // attributes out of this mft record.
2293        if (start < 0 || start >= allocated_size)
2294            ntfs_error(vol->sb, "Not enough space in the mft "
2295                    "record for the extended attribute "
2296                    "record. This case is not "
2297                    "implemented yet.");
2298        err = -EOPNOTSUPP;
2299        goto undo_alloc;
2300    }
2301    mp_rebuilt = true;
2302    /* Generate the mapping pairs array directly into the attr record. */
2303    err = ntfs_mapping_pairs_build(vol, (u8*)a +
2304            le16_to_cpu(a->data.non_resident.mapping_pairs_offset),
2305            mp_size, rl2, ll, -1, NULL);
2306    if (unlikely(err)) {
2307        if (start < 0 || start >= allocated_size)
2308            ntfs_error(vol->sb, "Cannot extend allocation of "
2309                    "inode 0x%lx, attribute type 0x%x, "
2310                    "because building the mapping pairs "
2311                    "failed with error code %i.", vi->i_ino,
2312                    (unsigned)le32_to_cpu(ni->type), err);
2313        err = -EIO;
2314        goto undo_alloc;
2315    }
2316    /* Update the highest_vcn. */
2317    a->data.non_resident.highest_vcn = cpu_to_sle64((new_alloc_size >>
2318            vol->cluster_size_bits) - 1);
2319    /*
2320     * We now have extended the allocated size of the attribute. Reflect
2321     * this in the ntfs_inode structure and the attribute record.
2322     */
2323    if (a->data.non_resident.lowest_vcn) {
2324        /*
2325         * We are not in the first attribute extent, switch to it, but
2326         * first ensure the changes will make it to disk later.
2327         */
2328        flush_dcache_mft_record_page(ctx->ntfs_ino);
2329        mark_mft_record_dirty(ctx->ntfs_ino);
2330        ntfs_attr_reinit_search_ctx(ctx);
2331        err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
2332                CASE_SENSITIVE, 0, NULL, 0, ctx);
2333        if (unlikely(err))
2334            goto restore_undo_alloc;
2335        /* @m is not used any more so no need to set it. */
2336        a = ctx->attr;
2337    }
2338    write_lock_irqsave(&ni->size_lock, flags);
2339    ni->allocated_size = new_alloc_size;
2340    a->data.non_resident.allocated_size = cpu_to_sle64(new_alloc_size);
2341    /*
2342     * FIXME: This would fail if @ni is a directory, $MFT, or an index,
2343     * since those can have sparse/compressed set. For example can be
2344     * set compressed even though it is not compressed itself and in that
2345     * case the bit means that files are to be created compressed in the
2346     * directory... At present this is ok as this code is only called for
2347     * regular files, and only for their $DATA attribute(s).
2348     * FIXME: The calculation is wrong if we created a hole above. For now
2349     * it does not matter as we never create holes.
2350     */
2351    if (NInoSparse(ni) || NInoCompressed(ni)) {
2352        ni->itype.compressed.size += new_alloc_size - allocated_size;
2353        a->data.non_resident.compressed_size =
2354                cpu_to_sle64(ni->itype.compressed.size);
2355        vi->i_blocks = ni->itype.compressed.size >> 9;
2356    } else
2357        vi->i_blocks = new_alloc_size >> 9;
2358    write_unlock_irqrestore(&ni->size_lock, flags);
2359alloc_done:
2360    if (new_data_size >= 0) {
2361        BUG_ON(new_data_size <
2362                sle64_to_cpu(a->data.non_resident.data_size));
2363        a->data.non_resident.data_size = cpu_to_sle64(new_data_size);
2364    }
2365flush_done:
2366    /* Ensure the changes make it to disk. */
2367    flush_dcache_mft_record_page(ctx->ntfs_ino);
2368    mark_mft_record_dirty(ctx->ntfs_ino);
2369done:
2370    ntfs_attr_put_search_ctx(ctx);
2371    unmap_mft_record(base_ni);
2372    up_write(&ni->runlist.lock);
2373    ntfs_debug("Done, new_allocated_size 0x%llx.",
2374            (unsigned long long)new_alloc_size);
2375    return new_alloc_size;
2376restore_undo_alloc:
2377    if (start < 0 || start >= allocated_size)
2378        ntfs_error(vol->sb, "Cannot complete extension of allocation "
2379                "of inode 0x%lx, attribute type 0x%x, because "
2380                "lookup of first attribute extent failed with "
2381                "error code %i.", vi->i_ino,
2382                (unsigned)le32_to_cpu(ni->type), err);
2383    if (err == -ENOENT)
2384        err = -EIO;
2385    ntfs_attr_reinit_search_ctx(ctx);
2386    if (ntfs_attr_lookup(ni->type, ni->name, ni->name_len, CASE_SENSITIVE,
2387            allocated_size >> vol->cluster_size_bits, NULL, 0,
2388            ctx)) {
2389        ntfs_error(vol->sb, "Failed to find last attribute extent of "
2390                "attribute in error code path. Run chkdsk to "
2391                "recover.");
2392        write_lock_irqsave(&ni->size_lock, flags);
2393        ni->allocated_size = new_alloc_size;
2394        /*
2395         * FIXME: This would fail if @ni is a directory... See above.
2396         * FIXME: The calculation is wrong if we created a hole above.
2397         * For now it does not matter as we never create holes.
2398         */
2399        if (NInoSparse(ni) || NInoCompressed(ni)) {
2400            ni->itype.compressed.size += new_alloc_size -
2401                    allocated_size;
2402            vi->i_blocks = ni->itype.compressed.size >> 9;
2403        } else
2404            vi->i_blocks = new_alloc_size >> 9;
2405        write_unlock_irqrestore(&ni->size_lock, flags);
2406        ntfs_attr_put_search_ctx(ctx);
2407        unmap_mft_record(base_ni);
2408        up_write(&ni->runlist.lock);
2409        /*
2410         * The only thing that is now wrong is the allocated size of the
2411         * base attribute extent which chkdsk should be able to fix.
2412         */
2413        NVolSetErrors(vol);
2414        return err;
2415    }
2416    ctx->attr->data.non_resident.highest_vcn = cpu_to_sle64(
2417            (allocated_size >> vol->cluster_size_bits) - 1);
2418undo_alloc:
2419    ll = allocated_size >> vol->cluster_size_bits;
2420    if (ntfs_cluster_free(ni, ll, -1, ctx) < 0) {
2421        ntfs_error(vol->sb, "Failed to release allocated cluster(s) "
2422                "in error code path. Run chkdsk to recover "
2423                "the lost cluster(s).");
2424        NVolSetErrors(vol);
2425    }
2426    m = ctx->mrec;
2427    a = ctx->attr;
2428    /*
2429     * If the runlist truncation fails and/or the search context is no
2430     * longer valid, we cannot resize the attribute record or build the
2431     * mapping pairs array thus we mark the inode bad so that no access to
2432     * the freed clusters can happen.
2433     */
2434    if (ntfs_rl_truncate_nolock(vol, &ni->runlist, ll) || IS_ERR(m)) {
2435        ntfs_error(vol->sb, "Failed to %s in error code path. Run "
2436                "chkdsk to recover.", IS_ERR(m) ?
2437                "restore attribute search context" :
2438                "truncate attribute runlist");
2439        NVolSetErrors(vol);
2440    } else if (mp_rebuilt) {
2441        if (ntfs_attr_record_resize(m, a, attr_len)) {
2442            ntfs_error(vol->sb, "Failed to restore attribute "
2443                    "record in error code path. Run "
2444                    "chkdsk to recover.");
2445            NVolSetErrors(vol);
2446        } else /* if (success) */ {
2447            if (ntfs_mapping_pairs_build(vol, (u8*)a + le16_to_cpu(
2448                    a->data.non_resident.
2449                    mapping_pairs_offset), attr_len -
2450                    le16_to_cpu(a->data.non_resident.
2451                    mapping_pairs_offset), rl2, ll, -1,
2452                    NULL)) {
2453                ntfs_error(vol->sb, "Failed to restore "
2454                        "mapping pairs array in error "
2455                        "code path. Run chkdsk to "
2456                        "recover.");
2457                NVolSetErrors(vol);
2458            }
2459            flush_dcache_mft_record_page(ctx->ntfs_ino);
2460            mark_mft_record_dirty(ctx->ntfs_ino);
2461        }
2462    }
2463err_out:
2464    if (ctx)
2465        ntfs_attr_put_search_ctx(ctx);
2466    if (m)
2467        unmap_mft_record(base_ni);
2468    up_write(&ni->runlist.lock);
2469conv_err_out:
2470    ntfs_debug("Failed. Returning error code %i.", err);
2471    return err;
2472}
2473
2474/**
2475 * ntfs_attr_set - fill (a part of) an attribute with a byte
2476 * @ni: ntfs inode describing the attribute to fill
2477 * @ofs: offset inside the attribute at which to start to fill
2478 * @cnt: number of bytes to fill
2479 * @val: the unsigned 8-bit value with which to fill the attribute
2480 *
2481 * Fill @cnt bytes of the attribute described by the ntfs inode @ni starting at
2482 * byte offset @ofs inside the attribute with the constant byte @val.
2483 *
2484 * This function is effectively like memset() applied to an ntfs attribute.
2485 * Note thie function actually only operates on the page cache pages belonging
2486 * to the ntfs attribute and it marks them dirty after doing the memset().
2487 * Thus it relies on the vm dirty page write code paths to cause the modified
2488 * pages to be written to the mft record/disk.
2489 *
2490 * Return 0 on success and -errno on error. An error code of -ESPIPE means
2491 * that @ofs + @cnt were outside the end of the attribute and no write was
2492 * performed.
2493 */
2494int ntfs_attr_set(ntfs_inode *ni, const s64 ofs, const s64 cnt, const u8 val)
2495{
2496    ntfs_volume *vol = ni->vol;
2497    struct address_space *mapping;
2498    struct page *page;
2499    u8 *kaddr;
2500    pgoff_t idx, end;
2501    unsigned start_ofs, end_ofs, size;
2502
2503    ntfs_debug("Entering for ofs 0x%llx, cnt 0x%llx, val 0x%hx.",
2504            (long long)ofs, (long long)cnt, val);
2505    BUG_ON(ofs < 0);
2506    BUG_ON(cnt < 0);
2507    if (!cnt)
2508        goto done;
2509    /*
2510     * FIXME: Compressed and encrypted attributes are not supported when
2511     * writing and we should never have gotten here for them.
2512     */
2513    BUG_ON(NInoCompressed(ni));
2514    BUG_ON(NInoEncrypted(ni));
2515    mapping = VFS_I(ni)->i_mapping;
2516    /* Work out the starting index and page offset. */
2517    idx = ofs >> PAGE_CACHE_SHIFT;
2518    start_ofs = ofs & ~PAGE_CACHE_MASK;
2519    /* Work out the ending index and page offset. */
2520    end = ofs + cnt;
2521    end_ofs = end & ~PAGE_CACHE_MASK;
2522    /* If the end is outside the inode size return -ESPIPE. */
2523    if (unlikely(end > i_size_read(VFS_I(ni)))) {
2524        ntfs_error(vol->sb, "Request exceeds end of attribute.");
2525        return -ESPIPE;
2526    }
2527    end >>= PAGE_CACHE_SHIFT;
2528    /* If there is a first partial page, need to do it the slow way. */
2529    if (start_ofs) {
2530        page = read_mapping_page(mapping, idx, NULL);
2531        if (IS_ERR(page)) {
2532            ntfs_error(vol->sb, "Failed to read first partial "
2533                    "page (error, index 0x%lx).", idx);
2534            return PTR_ERR(page);
2535        }
2536        /*
2537         * If the last page is the same as the first page, need to
2538         * limit the write to the end offset.
2539         */
2540        size = PAGE_CACHE_SIZE;
2541        if (idx == end)
2542            size = end_ofs;
2543        kaddr = kmap_atomic(page, KM_USER0);
2544        memset(kaddr + start_ofs, val, size - start_ofs);
2545        flush_dcache_page(page);
2546        kunmap_atomic(kaddr, KM_USER0);
2547        set_page_dirty(page);
2548        page_cache_release(page);
2549        balance_dirty_pages_ratelimited(mapping);
2550        cond_resched();
2551        if (idx == end)
2552            goto done;
2553        idx++;
2554    }
2555    /* Do the whole pages the fast way. */
2556    for (; idx < end; idx++) {
2557        /* Find or create the current page. (The page is locked.) */
2558        page = grab_cache_page(mapping, idx);
2559        if (unlikely(!page)) {
2560            ntfs_error(vol->sb, "Insufficient memory to grab "
2561                    "page (index 0x%lx).", idx);
2562            return -ENOMEM;
2563        }
2564        kaddr = kmap_atomic(page, KM_USER0);
2565        memset(kaddr, val, PAGE_CACHE_SIZE);
2566        flush_dcache_page(page);
2567        kunmap_atomic(kaddr, KM_USER0);
2568        /*
2569         * If the page has buffers, mark them uptodate since buffer
2570         * state and not page state is definitive in 2.6 kernels.
2571         */
2572        if (page_has_buffers(page)) {
2573            struct buffer_head *bh, *head;
2574
2575            bh = head = page_buffers(page);
2576            do {
2577                set_buffer_uptodate(bh);
2578            } while ((bh = bh->b_this_page) != head);
2579        }
2580        /* Now that buffers are uptodate, set the page uptodate, too. */
2581        SetPageUptodate(page);
2582        /*
2583         * Set the page and all its buffers dirty and mark the inode
2584         * dirty, too. The VM will write the page later on.
2585         */
2586        set_page_dirty(page);
2587        /* Finally unlock and release the page. */
2588        unlock_page(page);
2589        page_cache_release(page);
2590        balance_dirty_pages_ratelimited(mapping);
2591        cond_resched();
2592    }
2593    /* If there is a last partial page, need to do it the slow way. */
2594    if (end_ofs) {
2595        page = read_mapping_page(mapping, idx, NULL);
2596        if (IS_ERR(page)) {
2597            ntfs_error(vol->sb, "Failed to read last partial page "
2598                    "(error, index 0x%lx).", idx);
2599            return PTR_ERR(page);
2600        }
2601        kaddr = kmap_atomic(page, KM_USER0);
2602        memset(kaddr, val, end_ofs);
2603        flush_dcache_page(page);
2604        kunmap_atomic(kaddr, KM_USER0);
2605        set_page_dirty(page);
2606        page_cache_release(page);
2607        balance_dirty_pages_ratelimited(mapping);
2608        cond_resched();
2609    }
2610done:
2611    ntfs_debug("Done.");
2612    return 0;
2613}
2614
2615#endif /* NTFS_RW */
2616

Archive Download this file



interactive