Commit 33b0a50c authored by Anton Altaparmakov's avatar Anton Altaparmakov

Merge cantab.net:/usr/src/tng-2.0.11 into cantab.net:/usr/src/tng-2.0.12

parents 59b1ec78 66b0ca10
......@@ -247,6 +247,9 @@ ChangeLog
Note, a technical ChangeLog aimed at kernel hackers is in fs/ntfs/ChangeLog.
2.0.12:
- Internal cleanups in address space operations made possible by the
changes introduced in the previous release.
2.0.11:
- Internal updates and cleanups introducing the first step towards
fake inode based attribute i/o.
......
......@@ -6,7 +6,7 @@ ToDo:
user open()s a file with i_size > s_maxbytes? Should read_inode()
truncate the visible i_size? Will the user just get -E2BIG (or
whatever) on open()? Or will (s)he be able to open() but lseek() and
read() will fail when s_maxbytes is reached? -> Investigate this!
read() will fail when s_maxbytes is reached? -> Investigate this.
- Implement/allow non-resident index bitmaps in dir.c::ntfs_readdir()
and then also consider initialized_size w.r.t. the bitmaps, etc.
- vcn_to_lcn() should somehow return the correct pointer within the
......@@ -17,14 +17,28 @@ ToDo:
- Consider if ntfs_file_read_compressed_block() shouldn't be coping
with initialized_size < data_size. I don't think it can happen but
it requires more careful consideration.
- CLEANUP: Modularise and reuse code in aops.c. At the moment we have
several copies of almost identicall functions and the functions are
quite big. Modularising them a bit, e.g. a-la get_block(), will make
them cleaner and make code reuse easier.
- CLEANUP: At the moment we have two copies of almost identical
functions in aops.c, can merge them once fake inode address space
based attribute i/o is further developed.
- CLEANUP: Modularising code in aops.c a bit, e.g. a-la get_block(),
will be cleaner and make code reuse easier.
- Enable NFS exporting of NTFS.
- Use iget5_locked() and friends instead of conventional iget().
- Use fake inodes for address space i/o.
2.0.12 - Initial cleanup of address space operations following 2.0.11 changes.
- Merge fs/ntfs/aops.c::end_buffer_read_mst_async() and
fs/ntfs/aops.c::end_buffer_read_file_async() into one function
fs/ntfs/aops.c::end_buffer_read_attr_async() using NInoMstProtected()
to determine whether to apply mst fixups or not.
- Above change allows merging fs/ntfs/aops.c::ntfs_file_read_block()
and fs/ntfs/aops.c::ntfs_mst_readpage() into one function
fs/ntfs/aops.c::ntfs_attr_read_block(). Also, create a tiny wrapper
fs/ntfs/aops.c::ntfs_mst_readpage() to transform the parameters from
the VFS readpage function prototype to the ntfs_attr_read_block()
function prototype.
2.0.11 - Initial preparations for fake inode based attribute i/o.
- Move definition of ntfs_inode_state_bits to fs/ntfs/inode.h and
......
......@@ -5,7 +5,7 @@ obj-$(CONFIG_NTFS_FS) += ntfs.o
ntfs-objs := aops.o attrib.o compress.o debug.o dir.o file.o inode.o mft.o \
mst.o namei.o super.o sysctl.o time.o unistr.o upcase.o
EXTRA_CFLAGS = -DNTFS_VERSION=\"2.0.11\"
EXTRA_CFLAGS = -DNTFS_VERSION=\"2.0.12\"
ifeq ($(CONFIG_NTFS_DEBUG),y)
EXTRA_CFLAGS += -DDEBUG
......
......@@ -3,7 +3,7 @@
* Part of the Linux-NTFS project.
*
* Copyright (c) 2001,2002 Anton Altaparmakov.
* Copyright (C) 2002 Richard Russon.
* Copyright (c) 2002 Richard Russon.
*
* This program/include file is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as published
......@@ -30,31 +30,43 @@
#include "ntfs.h"
/**
* end_buffer_read_file_async -
* end_buffer_read_attr_async - async io completion for reading attributes
* @bh: buffer head on which io is completed
* @uptodate: whether @bh is now uptodate or not
*
* Async io completion handler for accessing files. Adapted from
* end_buffer_read_mst_async().
* Asynchronous I/O completion handler for reading pages belonging to the
* attribute address space of an inode. The inodes can either be files or
* directories or they can be fake inodes describing some attribute.
*
* If NInoMstProtected(), perform the post read mst fixups when all IO on the
* page has been completed and mark the page uptodate or set the error bit on
* the page. To determine the size of the records that need fixing up, we cheat
* a little bit by setting the index_block_size in ntfs_inode to the ntfs
* record size, and index_block_size_bits, to the log(base 2) of the ntfs
* record size.
*/
static void end_buffer_read_file_async(struct buffer_head *bh, int uptodate)
static void end_buffer_read_attr_async(struct buffer_head *bh, int uptodate)
{
static spinlock_t page_uptodate_lock = SPIN_LOCK_UNLOCKED;
unsigned long flags;
struct buffer_head *tmp;
struct page *page;
ntfs_inode *ni;
if (uptodate)
if (likely(uptodate))
set_buffer_uptodate(bh);
else
clear_buffer_uptodate(bh);
page = bh->b_page;
ni = NTFS_I(page->mapping->host);
if (likely(uptodate)) {
s64 file_ofs;
ntfs_inode *ni = NTFS_I(page->mapping->host);
file_ofs = (page->index << PAGE_CACHE_SHIFT) + bh_offset(bh);
/* Check for the current buffer head overflowing. */
if (file_ofs + bh->b_size > ni->initialized_size) {
char *addr;
int ofs = 0;
......@@ -82,10 +94,47 @@ static void end_buffer_read_file_async(struct buffer_head *bh, int uptodate)
SetPageError(page);
tmp = tmp->b_this_page;
}
spin_unlock_irqrestore(&page_uptodate_lock, flags);
if (!PageError(page))
/*
* If none of the buffers had errors then we can set the page uptodate,
* but we first have to perform the post read mst fixups, if the
* attribute is mst protected, i.e. if NInoMstProteced(ni) is true.
*/
if (!NInoMstProtected(ni)) {
if (likely(!PageError(page)))
SetPageUptodate(page);
unlock_page(page);
return;
} else {
char *addr;
unsigned int i, recs, nr_err;
u32 rec_size;
rec_size = ni->_IDM(index_block_size);
recs = PAGE_CACHE_SIZE / rec_size;
addr = kmap_atomic(page, KM_BIO_SRC_IRQ);
for (i = nr_err = 0; i < recs; i++) {
if (likely(!post_read_mst_fixup((NTFS_RECORD*)(addr +
i * rec_size), rec_size)))
continue;
nr_err++;
ntfs_error(ni->vol->sb, "post_read_mst_fixup() failed, "
"corrupt %s record 0x%Lx. Run chkdsk.",
ni->mft_no ? "index" : "mft",
(long long)(((s64)page->index <<
PAGE_CACHE_SHIFT >>
ni->_IDM(index_block_size_bits)) + i));
}
flush_dcache_page(page);
kunmap_atomic(addr, KM_BIO_SRC_IRQ);
if (likely(!nr_err && recs))
SetPageUptodate(page);
else {
ntfs_error(ni->vol->sb, "Setting page error, index "
"0x%lx.", page->index);
SetPageError(page);
}
}
unlock_page(page);
return;
still_busy:
......@@ -94,11 +143,20 @@ static void end_buffer_read_file_async(struct buffer_head *bh, int uptodate)
}
/**
* ntfs_file_read_block -
* ntfs_attr_read_block - fill a @page of an address space with data
* @page: page cache page to fill with data
*
* NTFS version of block_read_full_page(). Adapted from ntfs_mst_readpage().
* Fill the page @page of the address space belonging to the @page->host inode.
* We read each buffer asynchronously and when all buffers are read in, our io
* completion handler end_buffer_read_attr_async(), if required, automatically
* applies the mst fixups to the page before finally marking it uptodate and
* unlocking it.
*
* Return 0 on success and -errno on error.
*
* Contains an adapted version of fs/buffer.c::block_read_full_page().
*/
static int ntfs_file_read_block(struct page *page)
static int ntfs_attr_read_block(struct page *page)
{
VCN vcn;
LCN lcn;
......@@ -119,7 +177,7 @@ static int ntfs_file_read_block(struct page *page)
if (!page_has_buffers(page))
create_empty_buffers(page, blocksize, 0);
bh = head = page_buffers(page);
if (!bh)
if (unlikely(!bh))
return -ENOMEM;
blocks = PAGE_CACHE_SIZE >> blocksize_bits;
......@@ -128,11 +186,9 @@ static int ntfs_file_read_block(struct page *page)
zblock = (ni->initialized_size + blocksize - 1) >> blocksize_bits;
#ifdef DEBUG
if (unlikely(!ni->mft_no)) {
ntfs_error(vol->sb, "NTFS: Attempt to access $MFT! This is a "
"very serious bug! Denying access...");
return -EACCES;
}
if (unlikely(!ni->run_list.rl && !ni->mft_no))
panic("NTFS: $MFT/$DATA run list has been unmapped! This is a "
"very serious bug! Cannot continue...");
#endif
/* Loop through all the buffers in the page. */
......@@ -211,7 +267,7 @@ static int ntfs_file_read_block(struct page *page)
for (i = 0; i < nr; i++) {
struct buffer_head *tbh = arr[i];
lock_buffer(tbh);
tbh->b_end_io = end_buffer_read_file_async;
tbh->b_end_io = end_buffer_read_attr_async;
set_buffer_async_read(tbh);
}
/* Finally, start i/o on the buffers. */
......@@ -220,7 +276,7 @@ static int ntfs_file_read_block(struct page *page)
return 0;
}
/* No i/o was scheduled on any of the buffers. */
if (!PageError(page))
if (likely(!PageError(page)))
SetPageUptodate(page);
else /* Signal synchronous i/o error. */
nr = -EIO;
......@@ -234,17 +290,17 @@ static int ntfs_file_read_block(struct page *page)
* @page: page cache page to fill with data
*
* For non-resident attributes, ntfs_file_readpage() fills the @page of the open
* file @file by calling the generic block_read_full_page() function provided by
* the kernel which in turn invokes our ntfs_file_get_block() callback in order
* to create and read in the buffers associated with the page asynchronously.
* file @file by calling the ntfs version of the generic block_read_full_page()
* function provided by the kernel, ntfs_attr_read_block(), which in turn
* creates and reads in the buffers associated with the page asynchronously.
*
* For resident attributes, OTOH, ntfs_file_readpage() fills @page by copying
* the data from the mft record (which at this stage is most likely in memory)
* and fills the remainder with zeroes. Thus, in this case I/O is synchronous,
* and fills the remainder with zeroes. Thus, in this case, I/O is synchronous,
* as even if the mft record is not cached at this point in time, we need to
* wait for it to be read in before we can do the copy.
*
* Return zero on success or -errno on error.
* Return 0 on success or -errno on error.
*/
static int ntfs_file_readpage(struct file *file, struct page *page)
{
......@@ -256,43 +312,43 @@ static int ntfs_file_readpage(struct file *file, struct page *page)
u32 attr_len;
int err = 0;
if (!PageLocked(page))
if (unlikely(!PageLocked(page)))
PAGE_BUG(page);
ni = NTFS_I(page->mapping->host);
/* Is the unnamed $DATA attribute resident? */
if (test_bit(NI_NonResident, &ni->state)) {
if (NInoNonResident(ni)) {
/* Attribute is not resident. */
/* If the file is encrypted, we deny access, just like NT4. */
if (test_bit(NI_Encrypted, &ni->state)) {
if (NInoEncrypted(ni)) {
err = -EACCES;
goto unl_err_out;
}
/* Compressed data stream. Handled in compress.c. */
if (test_bit(NI_Compressed, &ni->state))
if (NInoCompressed(ni))
return ntfs_file_read_compressed_block(page);
/* Normal data stream. */
return ntfs_file_read_block(page);
return ntfs_attr_read_block(page);
}
/* Attribute is resident, implying it is not compressed or encrypted. */
/* Map, pin and lock the mft record for reading. */
mrec = map_mft_record(READ, ni);
if (IS_ERR(mrec)) {
if (unlikely(IS_ERR(mrec))) {
err = PTR_ERR(mrec);
goto unl_err_out;
}
ctx = get_attr_search_ctx(ni, mrec);
if (!ctx) {
if (unlikely(!ctx)) {
err = -ENOMEM;
goto unm_unl_err_out;
}
/* Find the data attribute in the mft record. */
if (!lookup_attr(AT_DATA, NULL, 0, 0, 0, NULL, 0, ctx)) {
if (unlikely(!lookup_attr(AT_DATA, NULL, 0, 0, 0, NULL, 0, ctx))) {
err = -ENOENT;
goto put_unm_unl_err_out;
}
......@@ -330,6 +386,25 @@ static int ntfs_file_readpage(struct file *file, struct page *page)
return err;
}
/**
* ntfs_mst_readpage - fill a @page of the mft or a directory with data
* @file: open file/directory to which the @page belongs or NULL
* @page: page cache page to fill with data
*
* Readpage method for the VFS address space operations of directory inodes
* and the $MFT/$DATA attribute.
*
* We just call ntfs_attr_read_block() here, in fact we only need this wrapper
* because of the difference in function parameters.
*/
int ntfs_mst_readpage(struct file *file, struct page *page)
{
if (unlikely(!PageLocked(page)))
PAGE_BUG(page);
return ntfs_attr_read_block(page);
}
/**
* end_buffer_read_mftbmp_async -
*
......@@ -343,7 +418,7 @@ static void end_buffer_read_mftbmp_async(struct buffer_head *bh, int uptodate)
struct buffer_head *tmp;
struct page *page;
if (uptodate)
if (likely(uptodate))
set_buffer_uptodate(bh);
else
clear_buffer_uptodate(bh);
......@@ -386,7 +461,7 @@ static void end_buffer_read_mftbmp_async(struct buffer_head *bh, int uptodate)
}
spin_unlock_irqrestore(&page_uptodate_lock, flags);
if (!PageError(page))
if (likely(!PageError(page)))
SetPageUptodate(page);
unlock_page(page);
return;
......@@ -410,7 +485,7 @@ static int ntfs_mftbmp_readpage(ntfs_volume *vol, struct page *page)
int nr, i;
unsigned char blocksize_bits;
if (!PageLocked(page))
if (unlikely(!PageLocked(page)))
PAGE_BUG(page);
blocksize = vol->sb->s_blocksize;
......@@ -419,7 +494,7 @@ static int ntfs_mftbmp_readpage(ntfs_volume *vol, struct page *page)
if (!page_has_buffers(page))
create_empty_buffers(page, blocksize, 0);
bh = head = page_buffers(page);
if (!bh)
if (unlikely(!bh))
return -ENOMEM;
blocks = PAGE_CACHE_SIZE >> blocksize_bits;
......@@ -503,264 +578,7 @@ static int ntfs_mftbmp_readpage(ntfs_volume *vol, struct page *page)
return 0;
}
/* No i/o was scheduled on any of the buffers. */
if (!PageError(page))
SetPageUptodate(page);
else /* Signal synchronous i/o error. */
nr = -EIO;
unlock_page(page);
return nr;
}
/**
* end_buffer_read_mst_async - async io completion for reading index records
* @bh: buffer head on which io is completed
* @uptodate: whether @bh is now uptodate or not
*
* Asynchronous I/O completion handler for reading pages belonging to the
* index allocation attribute address space of directory inodes.
*
* Perform the post read mst fixups when all IO on the page has been completed
* and marks the page uptodate or sets the error bit on the page.
*
* Adapted from fs/buffer.c.
*
* NOTE: We use this function as async io completion handler for reading pages
* belonging to the mft data attribute address space, too as this saves
* duplicating an almost identical function. We do this by cheating a little
* bit in setting the index_block_size in the mft ntfs_inode to the mft record
* size of the volume (vol->mft_record_size), and index_block_size_bits to
* mft_record_size_bits, respectively.
*/
static void end_buffer_read_mst_async(struct buffer_head *bh, int uptodate)
{
static spinlock_t page_uptodate_lock = SPIN_LOCK_UNLOCKED;
unsigned long flags;
struct buffer_head *tmp;
struct page *page;
ntfs_inode *ni;
if (uptodate)
set_buffer_uptodate(bh);
else
clear_buffer_uptodate(bh);
page = bh->b_page;
ni = NTFS_I(page->mapping->host);
if (likely(uptodate)) {
s64 file_ofs;
file_ofs = (page->index << PAGE_CACHE_SHIFT) + bh_offset(bh);
/* Check for the current buffer head overflowing. */
if (file_ofs + bh->b_size > ni->initialized_size) {
char *addr;
int ofs = 0;
if (file_ofs < ni->initialized_size)
ofs = ni->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);
kunmap_atomic(addr, KM_BIO_SRC_IRQ);
}
} else
SetPageError(page);
spin_lock_irqsave(&page_uptodate_lock, flags);
clear_buffer_async_read(bh);
unlock_buffer(bh);
tmp = bh->b_this_page;
while (tmp != bh) {
if (buffer_locked(tmp)) {
if (buffer_async_read(tmp))
goto still_busy;
} else if (!buffer_uptodate(tmp))
SetPageError(page);
tmp = tmp->b_this_page;
}
spin_unlock_irqrestore(&page_uptodate_lock, flags);
/*
* If none of the buffers had errors then we can set the page uptodate,
* but we first have to perform the post read mst fixups.
*/
if (!PageError(page)) {
char *addr;
unsigned int i, recs, nr_err = 0;
u32 rec_size;
rec_size = ni->_IDM(index_block_size);
recs = PAGE_CACHE_SIZE / rec_size;
addr = kmap_atomic(page, KM_BIO_SRC_IRQ);
for (i = 0; i < recs; i++) {
if (!post_read_mst_fixup((NTFS_RECORD*)(addr +
i * rec_size), rec_size))
continue;
nr_err++;
ntfs_error(ni->vol->sb, "post_read_mst_fixup() failed, "
"corrupt %s record 0x%Lx. Run chkdsk.",
ni->mft_no ? "index" : "mft",
(long long)((page->index <<
PAGE_CACHE_SHIFT >>
ni->_IDM(index_block_size_bits)) + i));
}
flush_dcache_page(page);
kunmap_atomic(addr, KM_BIO_SRC_IRQ);
if (likely(!nr_err && recs))
SetPageUptodate(page);
else {
ntfs_error(ni->vol->sb, "Setting page error, index "
"0x%lx.", page->index);
SetPageError(page);
}
}
unlock_page(page);
return;
still_busy:
spin_unlock_irqrestore(&page_uptodate_lock, flags);
return;
}
/**
* ntfs_mst_readpage - fill a @page of the mft or a directory with data
* @file: open file/directory to which the page @page belongs or NULL
* @page: page cache page to fill with data
*
* Readpage method for the VFS address space operations.
*
* Fill the page @page of the $MFT or the open directory @dir. We read each
* buffer asynchronously and when all buffers are read in our io completion
* handler end_buffer_read_mst_async() automatically applies the mst fixups to
* the page before finally marking it uptodate and unlocking it.
*
* Contains an adapted version of fs/buffer.c::block_read_full_page().
*/
int ntfs_mst_readpage(struct file *dir, struct page *page)
{
VCN vcn;
LCN lcn;
ntfs_inode *ni;
ntfs_volume *vol;
struct buffer_head *bh, *head, *arr[MAX_BUF_PER_PAGE];
sector_t iblock, lblock, zblock;
unsigned int blocksize, blocks, vcn_ofs;
int i, nr;
unsigned char blocksize_bits;
if (!PageLocked(page))
PAGE_BUG(page);
ni = NTFS_I(page->mapping->host);
vol = ni->vol;
blocksize_bits = VFS_I(ni)->i_blkbits;
blocksize = 1 << blocksize_bits;
if (!page_has_buffers(page))
create_empty_buffers(page, blocksize, 0);
bh = head = page_buffers(page);
if (!bh)
return -ENOMEM;
blocks = PAGE_CACHE_SIZE >> blocksize_bits;
iblock = page->index << (PAGE_CACHE_SHIFT - blocksize_bits);
lblock = (ni->allocated_size + blocksize - 1) >> blocksize_bits;
zblock = (ni->initialized_size + blocksize - 1) >> blocksize_bits;
#ifdef DEBUG
if (unlikely(!ni->run_list.rl && !ni->mft_no))
panic("NTFS: $MFT/$DATA run list has been unmapped! This is a "
"very serious bug! Cannot continue...");
#endif
/* Loop through all the buffers in the page. */
nr = i = 0;
do {
if (unlikely(buffer_uptodate(bh)))
continue;
if (unlikely(buffer_mapped(bh))) {
arr[nr++] = bh;
continue;
}
bh->b_bdev = vol->sb->s_bdev;
/* Is the block within the allowed limits? */
if (iblock < lblock) {
BOOL is_retry = FALSE;
/* Convert iblock into corresponding vcn and offset. */
vcn = (VCN)iblock << blocksize_bits >>
vol->cluster_size_bits;
vcn_ofs = ((VCN)iblock << blocksize_bits) &
vol->cluster_size_mask;
retry_remap:
/* Convert the vcn to the corresponding lcn. */
down_read(&ni->run_list.lock);
lcn = vcn_to_lcn(ni->run_list.rl, vcn);
up_read(&ni->run_list.lock);
/* Successful remap. */
if (lcn >= 0) {
/* Setup buffer head to correct block. */
bh->b_blocknr = ((lcn << vol->cluster_size_bits)
+ vcn_ofs) >> blocksize_bits;
set_buffer_mapped(bh);
/* Only read initialized data blocks. */
if (iblock < zblock) {
arr[nr++] = bh;
continue;
}
/* Fully non-initialized data block, zero it. */
goto handle_zblock;
}
/* It is a hole, need to zero it. */
if (lcn == LCN_HOLE)
goto handle_hole;
/* If first try and run list unmapped, map and retry. */
if (!is_retry && lcn == LCN_RL_NOT_MAPPED) {
is_retry = TRUE;
if (!map_run_list(ni, vcn))
goto retry_remap;
}
/* Hard error, zero out region. */
SetPageError(page);
ntfs_error(vol->sb, "vcn_to_lcn(vcn = 0x%Lx) failed "
"with error code 0x%Lx%s.",
(long long)vcn, (long long)-lcn,
is_retry ? " even after retrying" : "");
// FIXME: Depending on vol->on_errors, do something.
}
/*
* Either iblock was outside lblock limits or vcn_to_lcn()
* returned error. Just zero that portion of the page and set
* the buffer uptodate.
*/
handle_hole:
bh->b_blocknr = -1UL;
clear_buffer_mapped(bh);
handle_zblock:
memset(kmap(page) + i * blocksize, 0, blocksize);
flush_dcache_page(page);
kunmap(page);
set_buffer_uptodate(bh);
} while (i++, iblock++, (bh = bh->b_this_page) != head);
/* Check we have at least one buffer ready for i/o. */
if (nr) {
/* Lock the buffers. */
for (i = 0; i < nr; i++) {
struct buffer_head *tbh = arr[i];
lock_buffer(tbh);
tbh->b_end_io = end_buffer_read_mst_async;
set_buffer_async_read(tbh);
}
/* Finally, start i/o on the buffers. */
for (i = 0; i < nr; i++)
submit_bh(READ, arr[i]);
return 0;
}
/* No i/o was scheduled on any of the buffers. */
if (!PageError(page))
if (likely(!PageError(page)))
SetPageUptodate(page);
else /* Signal synchronous i/o error. */
nr = -EIO;
......
......@@ -102,7 +102,7 @@ extern int ntfs_mst_readpage(struct file *, struct page *);
* ntfs_mft_aops - address space operations for access to $MFT
*
* Address space operations for access to $MFT. This allows us to simply use
* read_cache_page() in map_mft_record().
* ntfs_map_page() in map_mft_record_page().
*/
struct address_space_operations ntfs_mft_aops = {
writepage: NULL, /* Write dirty page to disk. */
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment