* aops.c - NTFS kernel address space operations and page cache handling.
* Part of the Linux-NTFS project.
*
- * Copyright (c) 2001-2004 Anton Altaparmakov
+ * Copyright (c) 2001-2005 Anton Altaparmakov
* Copyright (c) 2002 Richard Russon
*
* This program/include file is free software; you can redistribute it and/or
ni = NTFS_I(page->mapping->host);
if (likely(uptodate)) {
- s64 file_ofs;
+ s64 file_ofs, initialized_size;
set_buffer_uptodate(bh);
file_ofs = ((s64)page->index << PAGE_CACHE_SHIFT) +
bh_offset(bh);
+ read_lock_irqsave(&ni->size_lock, flags);
+ initialized_size = ni->initialized_size;
+ read_unlock_irqrestore(&ni->size_lock, flags);
/* Check for the current buffer head overflowing. */
- if (file_ofs + bh->b_size > ni->initialized_size) {
+ if (file_ofs + bh->b_size > initialized_size) {
char *addr;
int ofs = 0;
- if (file_ofs < ni->initialized_size)
- ofs = ni->initialized_size - file_ofs;
+ if (file_ofs < initialized_size)
+ ofs = initialized_size - file_ofs;
addr = kmap_atomic(page, KM_BIO_SRC_IRQ);
memset(addr + bh_offset(bh) + ofs, 0, bh->b_size - ofs);
flush_dcache_page(page);
i * rec_size), rec_size);
flush_dcache_page(page);
kunmap_atomic(addr, KM_BIO_SRC_IRQ);
- if (likely(!PageError(page) && page_uptodate))
+ if (likely(page_uptodate && !PageError(page)))
SetPageUptodate(page);
}
unlock_page(page);
runlist_element *rl;
struct buffer_head *bh, *head, *arr[MAX_BUF_PER_PAGE];
sector_t iblock, lblock, zblock;
+ unsigned long flags;
unsigned int blocksize, vcn_ofs;
int i, nr;
unsigned char blocksize_bits;
}
iblock = (s64)page->index << (PAGE_CACHE_SHIFT - blocksize_bits);
+ read_lock_irqsave(&ni->size_lock, flags);
lblock = (ni->allocated_size + blocksize - 1) >> blocksize_bits;
zblock = (ni->initialized_size + blocksize - 1) >> blocksize_bits;
+ read_unlock_irqrestore(&ni->size_lock, flags);
/* Loop through all the buffers in the page. */
rl = NULL;
goto lock_retry_remap;
rl = NULL;
lcn = err;
- }
+ } else if (!rl)
+ up_read(&ni->runlist.lock);
/* Hard error, zero out region. */
bh->b_blocknr = -1;
SetPageError(page);
*/
static int ntfs_readpage(struct file *file, struct page *page)
{
- loff_t i_size;
ntfs_inode *ni, *base_ni;
u8 *kaddr;
ntfs_attr_search_ctx *ctx;
MFT_RECORD *mrec;
+ unsigned long flags;
u32 attr_len;
int err = 0;
+retry_readpage:
BUG_ON(!PageLocked(page));
/*
* This can potentially happen because we clear PageUptodate() during
* Attribute is resident, implying it is not compressed or encrypted.
* This also means the attribute is smaller than an mft record and
* hence smaller than a page, so can simply zero out any pages with
- * index above 0. We can also do this if the file size is 0.
+ * index above 0.
*/
- if (unlikely(page->index > 0 || !i_size_read(VFS_I(ni)))) {
+ if (unlikely(page->index > 0)) {
kaddr = kmap_atomic(page, KM_USER0);
memset(kaddr, 0, PAGE_CACHE_SIZE);
flush_dcache_page(page);
err = PTR_ERR(mrec);
goto err_out;
}
+ /*
+ * If a parallel write made the attribute non-resident, drop the mft
+ * record and retry the readpage.
+ */
+ if (unlikely(NInoNonResident(ni))) {
+ unmap_mft_record(base_ni);
+ goto retry_readpage;
+ }
ctx = ntfs_attr_get_search_ctx(base_ni, mrec);
if (unlikely(!ctx)) {
err = -ENOMEM;
if (unlikely(err))
goto put_unm_err_out;
attr_len = le32_to_cpu(ctx->attr->data.resident.value_length);
- i_size = i_size_read(VFS_I(ni));
- if (unlikely(attr_len > i_size))
- attr_len = i_size;
+ read_lock_irqsave(&ni->size_lock, flags);
+ if (unlikely(attr_len > ni->initialized_size))
+ attr_len = ni->initialized_size;
+ read_unlock_irqrestore(&ni->size_lock, flags);
kaddr = kmap_atomic(page, KM_USER0);
/* Copy the data to the page. */
memcpy(kaddr, (u8*)ctx->attr +
{
VCN vcn;
LCN lcn;
+ s64 initialized_size;
+ loff_t i_size;
sector_t block, dblock, iblock;
struct inode *vi;
ntfs_inode *ni;
ntfs_volume *vol;
runlist_element *rl;
struct buffer_head *bh, *head;
+ unsigned long flags;
unsigned int blocksize, vcn_ofs;
int err;
BOOL need_end_writeback;
/* The first block in the page. */
block = (s64)page->index << (PAGE_CACHE_SHIFT - blocksize_bits);
+ read_lock_irqsave(&ni->size_lock, flags);
+ i_size = i_size_read(vi);
+ initialized_size = ni->initialized_size;
+ read_unlock_irqrestore(&ni->size_lock, flags);
+
/* The first out of bounds block for the data size. */
- dblock = (vi->i_size + blocksize - 1) >> blocksize_bits;
+ dblock = (i_size + blocksize - 1) >> blocksize_bits;
/* The last (fully or partially) initialized block. */
- iblock = ni->initialized_size >> blocksize_bits;
+ iblock = initialized_size >> blocksize_bits;
/*
* Be very careful. We have no exclusion from __set_page_dirty_buffers
/* Make sure we have enough initialized size. */
if (unlikely((block >= iblock) &&
- (ni->initialized_size < vi->i_size))) {
+ (initialized_size < i_size))) {
/*
* If this page is fully outside initialized size, zero
* out all pages between the current initialized size
}
/* It is a hole, need to instantiate it. */
if (lcn == LCN_HOLE) {
+ u8 *kaddr;
+ unsigned long *bpos, *bend;
+
+ /* Check if the buffer is zero. */
+ kaddr = kmap_atomic(page, KM_USER0);
+ bpos = (unsigned long *)(kaddr + bh_offset(bh));
+ bend = (unsigned long *)((u8*)bpos + blocksize);
+ do {
+ if (unlikely(*bpos))
+ break;
+ } while (likely(++bpos < bend));
+ kunmap_atomic(kaddr, KM_USER0);
+ if (bpos == bend) {
+ /*
+ * Buffer is zero and sparse, no need to write
+ * it.
+ */
+ bh->b_blocknr = -1;
+ clear_buffer_dirty(bh);
+ continue;
+ }
// TODO: Instantiate the hole.
// clear_buffer_new(bh);
// unmap_underlying_metadata(bh->b_bdev, bh->b_blocknr);
goto lock_retry_remap;
rl = NULL;
lcn = err;
- }
+ } else if (!rl)
+ up_read(&ni->runlist.lock);
/* Failed to map the buffer, even after retrying. */
bh->b_blocknr = -1;
ntfs_error(vol->sb, "Failed to write to inode 0x%lx, "
/* For the error case, need to reset bh to the beginning. */
bh = head;
- /* Just an optimization, so ->readpage() isn't called later. */
+ /* Just an optimization, so ->readpage() is not called later. */
if (unlikely(!PageUptodate(page))) {
int uptodate = 1;
do {
/* Setup all mapped, dirty buffers for async write i/o. */
do {
- get_bh(bh);
if (buffer_mapped(bh) && buffer_dirty(bh)) {
lock_buffer(bh);
if (test_clear_buffer_dirty(bh)) {
BUG_ON(PageWriteback(page));
set_page_writeback(page); /* Keeps try_to_free_buffers() away. */
- unlock_page(page);
- /*
- * Submit the prepared buffers for i/o. Note the page is unlocked,
- * and the async write i/o completion handler can end_page_writeback()
- * at any time after the *first* submit_bh(). So the buffers can then
- * disappear...
- */
+ /* Submit the prepared buffers for i/o. */
need_end_writeback = TRUE;
do {
struct buffer_head *next = bh->b_this_page;
submit_bh(WRITE, bh);
need_end_writeback = FALSE;
}
- put_bh(bh);
bh = next;
} while (bh != head);
+ unlock_page(page);
/* If no i/o was started, need to end_page_writeback(). */
if (unlikely(need_end_writeback))
ntfs_inode *ni = NTFS_I(vi);
ntfs_volume *vol = ni->vol;
u8 *kaddr;
- unsigned char bh_size_bits = vi->i_blkbits;
- unsigned int bh_size = 1 << bh_size_bits;
unsigned int rec_size = ni->itype.index.block_size;
ntfs_inode *locked_nis[PAGE_CACHE_SIZE / rec_size];
struct buffer_head *bh, *head, *tbh, *rec_start_bh;
- int max_bhs = PAGE_CACHE_SIZE / bh_size;
- struct buffer_head *bhs[max_bhs];
+ struct buffer_head *bhs[MAX_BUF_PER_PAGE];
runlist_element *rl;
- int i, nr_locked_nis, nr_recs, nr_bhs, bhs_per_rec, err, err2;
- unsigned rec_size_bits;
+ int i, nr_locked_nis, nr_recs, nr_bhs, max_bhs, bhs_per_rec, err, err2;
+ unsigned bh_size, rec_size_bits;
BOOL sync, is_mft, page_is_dirty, rec_is_dirty;
+ unsigned char bh_size_bits;
ntfs_debug("Entering for inode 0x%lx, attribute type 0x%x, page index "
"0x%lx.", vi->i_ino, ni->type, page->index);
*/
BUG_ON(!(is_mft || S_ISDIR(vi->i_mode) ||
(NInoAttr(ni) && ni->type == AT_INDEX_ALLOCATION)));
+ bh_size_bits = vi->i_blkbits;
+ bh_size = 1 << bh_size_bits;
+ max_bhs = PAGE_CACHE_SIZE / bh_size;
BUG_ON(!max_bhs);
+ BUG_ON(max_bhs > MAX_BUF_PER_PAGE);
/* Were we called for sync purposes? */
sync = (wbc->sync_mode == WB_SYNC_ALL);
(PAGE_CACHE_SHIFT - bh_size_bits);
/* The first out of bounds block for the data size. */
- dblock = (vi->i_size + bh_size - 1) >> bh_size_bits;
+ dblock = (i_size_read(vi) + bh_size - 1) >> bh_size_bits;
rl = NULL;
err = err2 = nr_bhs = nr_recs = nr_locked_nis = 0;
if (likely(block < rec_block)) {
if (unlikely(block >= dblock)) {
clear_buffer_dirty(bh);
+ set_buffer_uptodate(bh);
continue;
}
/*
LCN lcn;
unsigned int vcn_ofs;
+ bh->b_bdev = vol->sb->s_bdev;
/* Obtain the vcn and offset of the current block. */
vcn = (VCN)block << bh_size_bits;
vcn_ofs = vcn & vol->cluster_size_mask;
if (err2 == -ENOMEM)
page_is_dirty = TRUE;
lcn = err2;
- } else
+ } else {
err2 = -EIO;
+ if (!rl)
+ up_read(&ni->runlist.lock);
+ }
/* Hard error. Abort writing this record. */
if (!err || err == -ENOMEM)
err = err2;
"attribute type 0x%x) because "
"its location on disk could "
"not be determined (error "
- "code %lli).", (s64)block <<
+ "code %lli).",
+ (long long)block <<
bh_size_bits >>
vol->mft_record_size_bits,
ni->mft_no, ni->type,
static int ntfs_writepage(struct page *page, struct writeback_control *wbc)
{
loff_t i_size;
- struct inode *vi;
- ntfs_inode *ni, *base_ni;
+ struct inode *vi = page->mapping->host;
+ ntfs_inode *base_ni = NULL, *ni = NTFS_I(vi);
char *kaddr;
- ntfs_attr_search_ctx *ctx;
- MFT_RECORD *m;
+ ntfs_attr_search_ctx *ctx = NULL;
+ MFT_RECORD *m = NULL;
u32 attr_len;
int err;
+retry_writepage:
BUG_ON(!PageLocked(page));
-
- vi = page->mapping->host;
i_size = i_size_read(vi);
-
/* Is the page fully outside i_size? (truncate in progress) */
if (unlikely(page->index >= (i_size + PAGE_CACHE_SIZE - 1) >>
PAGE_CACHE_SHIFT)) {
ntfs_debug("Write outside i_size - truncated?");
return 0;
}
- ni = NTFS_I(vi);
-
+ /*
+ * Only $DATA attributes can be encrypted and only unnamed $DATA
+ * attributes can be compressed. Index root can have the flags set but
+ * this means to create compressed/encrypted files, not that the
+ * attribute is compressed/encrypted.
+ */
+ if (ni->type != AT_INDEX_ROOT) {
+ /* If file is encrypted, deny access, just like NT4. */
+ if (NInoEncrypted(ni)) {
+ unlock_page(page);
+ BUG_ON(ni->type != AT_DATA);
+ ntfs_debug("Denying write access to encrypted "
+ "file.");
+ return -EACCES;
+ }
+ /* Compressed data streams are handled in compress.c. */
+ if (NInoNonResident(ni) && NInoCompressed(ni)) {
+ BUG_ON(ni->type != AT_DATA);
+ BUG_ON(ni->name_len);
+ // TODO: Implement and replace this with
+ // return ntfs_write_compressed_block(page);
+ unlock_page(page);
+ ntfs_error(vi->i_sb, "Writing to compressed files is "
+ "not supported yet. Sorry.");
+ return -EOPNOTSUPP;
+ }
+ // TODO: Implement and remove this check.
+ if (NInoNonResident(ni) && NInoSparse(ni)) {
+ unlock_page(page);
+ ntfs_error(vi->i_sb, "Writing to sparse files is not "
+ "supported yet. Sorry.");
+ return -EOPNOTSUPP;
+ }
+ }
/* NInoNonResident() == NInoIndexAllocPresent() */
if (NInoNonResident(ni)) {
- /*
- * Only unnamed $DATA attributes can be compressed, encrypted,
- * and/or sparse.
- */
- if (ni->type == AT_DATA && !ni->name_len) {
- /* If file is encrypted, deny access, just like NT4. */
- if (NInoEncrypted(ni)) {
- unlock_page(page);
- ntfs_debug("Denying write access to encrypted "
- "file.");
- return -EACCES;
- }
- /* Compressed data streams are handled in compress.c. */
- if (NInoCompressed(ni)) {
- // TODO: Implement and replace this check with
- // return ntfs_write_compressed_block(page);
- unlock_page(page);
- ntfs_error(vi->i_sb, "Writing to compressed "
- "files is not supported yet. "
- "Sorry.");
- return -EOPNOTSUPP;
- }
- // TODO: Implement and remove this check.
- if (NInoSparse(ni)) {
- unlock_page(page);
- ntfs_error(vi->i_sb, "Writing to sparse files "
- "is not supported yet. Sorry.");
- return -EOPNOTSUPP;
- }
- }
/* We have to zero every time due to mmap-at-end-of-file. */
if (page->index >= (i_size >> PAGE_CACHE_SHIFT)) {
/* The page straddles i_size. */
/* Handle mst protected attributes. */
if (NInoMstProtected(ni))
return ntfs_write_mst_block(page, wbc);
- /* Normal data stream. */
+ /* Normal, non-resident data stream. */
return ntfs_write_block(page, wbc);
}
/*
- * Attribute is resident, implying it is not compressed, encrypted,
- * sparse, or mst protected. This also means the attribute is smaller
- * than an mft record and hence smaller than a page, so can simply
- * return error on any pages with index above 0.
+ * Attribute is resident, implying it is not compressed, encrypted, or
+ * mst protected. This also means the attribute is smaller than an mft
+ * record and hence smaller than a page, so can simply return error on
+ * any pages with index above 0. Note the attribute can actually be
+ * marked compressed but if it is resident the actual data is not
+ * compressed so we are ok to ignore the compressed flag here.
*/
BUG_ON(page_has_buffers(page));
BUG_ON(!PageUptodate(page));
ctx = NULL;
goto err_out;
}
+ /*
+ * If a parallel write made the attribute non-resident, drop the mft
+ * record and retry the writepage.
+ */
+ if (unlikely(NInoNonResident(ni))) {
+ unmap_mft_record(base_ni);
+ goto retry_writepage;
+ }
ctx = ntfs_attr_get_search_ctx(base_ni, m);
if (unlikely(!ctx)) {
err = -ENOMEM;
BUG_ON(PageWriteback(page));
set_page_writeback(page);
unlock_page(page);
-
/*
- * Here, we don't need to zero the out of bounds area everytime because
- * the below memcpy() already takes care of the mmap-at-end-of-file
- * requirements. If the file is converted to a non-resident one, then
- * the code path use is switched to the non-resident one where the
- * zeroing happens on each ntfs_writepage() invocation.
- *
- * The above also applies nicely when i_size is decreased.
- *
- * When i_size is increased, the memory between the old and new i_size
- * _must_ be zeroed (or overwritten with new data). Otherwise we will
- * expose data to userspace/disk which should never have been exposed.
- *
- * FIXME: Ensure that i_size increases do the zeroing/overwriting and
- * if we cannot guarantee that, then enable the zeroing below. If the
- * zeroing below is enabled, we MUST move the unlock_page() from above
- * to after the kunmap_atomic(), i.e. just before the
- * end_page_writeback().
- * UPDATE: ntfs_prepare/commit_write() do the zeroing on i_size
- * increases for resident attributes so those are ok.
- * TODO: ntfs_truncate(), others?
+ * Here, we do not need to zero the out of bounds area everytime
+ * because the below memcpy() already takes care of the
+ * mmap-at-end-of-file requirements. If the file is converted to a
+ * non-resident one, then the code path use is switched to the
+ * non-resident one where the zeroing happens on each ntfs_writepage()
+ * invocation.
*/
-
attr_len = le32_to_cpu(ctx->attr->data.resident.value_length);
- i_size = i_size_read(VFS_I(ni));
- kaddr = kmap_atomic(page, KM_USER0);
+ i_size = i_size_read(vi);
if (unlikely(attr_len > i_size)) {
- /* Zero out of bounds area in the mft record. */
- memset((u8*)ctx->attr + le16_to_cpu(
- ctx->attr->data.resident.value_offset) +
- i_size, 0, attr_len - i_size);
attr_len = i_size;
+ ctx->attr->data.resident.value_length = cpu_to_le32(attr_len);
}
+ kaddr = kmap_atomic(page, KM_USER0);
/* Copy the data from the page to the mft record. */
memcpy((u8*)ctx->attr +
le16_to_cpu(ctx->attr->data.resident.value_offset),
err = 0;
} else {
ntfs_error(vi->i_sb, "Resident attribute write failed with "
- "error %i. Setting page error flag.", err);
+ "error %i.", err);
SetPageError(page);
+ NVolSetErrors(ni->vol);
+ make_bad_inode(vi);
}
unlock_page(page);
if (ctx)
{
VCN vcn;
LCN lcn;
+ s64 initialized_size;
+ loff_t i_size;
sector_t block, ablock, iblock;
struct inode *vi;
ntfs_inode *ni;
ntfs_volume *vol;
runlist_element *rl;
struct buffer_head *bh, *head, *wait[2], **wait_bh = wait;
+ unsigned long flags;
unsigned int vcn_ofs, block_start, block_end, blocksize;
int err;
BOOL is_retry;
/* The first block in the page. */
block = (s64)page->index << (PAGE_CACHE_SHIFT - blocksize_bits);
+ read_lock_irqsave(&ni->size_lock, flags);
/*
- * The first out of bounds block for the allocated size. No need to
+ * The first out of bounds block for the allocated size. No need to
* round up as allocated_size is in multiples of cluster size and the
* minimum cluster size is 512 bytes, which is equal to the smallest
* blocksize.
*/
ablock = ni->allocated_size >> blocksize_bits;
+ i_size = i_size_read(vi);
+ initialized_size = ni->initialized_size;
+ read_unlock_irqrestore(&ni->size_lock, flags);
/* The last (fully or partially) initialized block. */
- iblock = ni->initialized_size >> blocksize_bits;
+ iblock = initialized_size >> blocksize_bits;
/* Loop through all the buffers in the page. */
block_start = 0;
* request, i.e. block < ablock is true.
*/
if (unlikely((block >= iblock) &&
- (ni->initialized_size < vi->i_size))) {
+ (initialized_size < i_size))) {
/*
* If this page is fully outside initialized size, zero
* out all pages between the current initialized size
"not supported yet. "
"Sorry.");
err = -EOPNOTSUPP;
+ if (!rl)
+ up_read(&ni->runlist.lock);
goto err_out;
} else if (!is_retry &&
lcn == LCN_RL_NOT_MAPPED) {
goto lock_retry_remap;
rl = NULL;
lcn = err;
- }
+ } else if (!rl)
+ up_read(&ni->runlist.lock);
/*
* Failed to map the buffer, even after
* retrying.
unsigned from, unsigned to)
{
s64 new_size;
+ loff_t i_size;
struct inode *vi = page->mapping->host;
ntfs_inode *base_ni = NULL, *ni = NTFS_I(vi);
ntfs_volume *vol = ni->vol;
BUG_ON(page_has_buffers(page));
new_size = ((s64)page->index << PAGE_CACHE_SHIFT) + to;
/* If we do not need to resize the attribute allocation we are done. */
- if (new_size <= vi->i_size)
+ if (new_size <= i_size_read(vi))
goto done;
-
- // FIXME: We abort for now as this code is not safe.
- ntfs_error(vi->i_sb, "Changing the file size is not supported yet. "
- "Sorry.");
- return -EOPNOTSUPP;
-
/* Map, pin, and lock the (base) mft record. */
if (!NInoAttr(ni))
base_ni = ni;
a = ctx->attr;
/* The total length of the attribute value. */
attr_len = le32_to_cpu(a->data.resident.value_length);
- BUG_ON(vi->i_size != attr_len);
+ /* Fix an eventual previous failure of ntfs_commit_write(). */
+ i_size = i_size_read(vi);
+ if (unlikely(attr_len > i_size)) {
+ attr_len = i_size;
+ a->data.resident.value_length = cpu_to_le32(attr_len);
+ }
+ /* If we do not need to resize the attribute allocation we are done. */
+ if (new_size <= attr_len)
+ goto done_unm;
/* Check if new size is allowed in $AttrDef. */
err = ntfs_attr_size_bounds_check(vol, ni->type, new_size);
if (unlikely(err)) {
}
flush_dcache_mft_record_page(ctx->ntfs_ino);
mark_mft_record_dirty(ctx->ntfs_ino);
+done_unm:
ntfs_attr_put_search_ctx(ctx);
unmap_mft_record(base_ni);
/*
* now we know ntfs_prepare_write() would have failed in the write
* exceeds i_size case, so this will never trigger which is fine.
*/
- if (pos > vi->i_size) {
+ if (pos > i_size_read(vi)) {
ntfs_error(vi->i_sb, "Writing beyond the existing file size is "
"not supported yet. Sorry.");
return -EOPNOTSUPP;
}
kunmap_atomic(kaddr, KM_USER0);
/* Update i_size if necessary. */
- if (vi->i_size < attr_len) {
+ if (i_size_read(vi) < attr_len) {
+ unsigned long flags;
+
+ write_lock_irqsave(&ni->size_lock, flags);
ni->allocated_size = ni->initialized_size = attr_len;
i_size_write(vi, attr_len);
+ write_unlock_irqrestore(&ni->size_lock, flags);
}
/* Mark the mft record dirty, so it gets written back. */
flush_dcache_mft_record_page(ctx->ntfs_ino);